No SIM slot. So I have to carry a mobile hotspot if I want to browse the internet outside my home. It’s like Apple forgets there is a world outside the US where eSIM is not a thing.
引用@hboonApple likes apps that show off new hardware. How about an iPhone Duo game where the primary mechanic is opening and closing it, say to squash a fly? ↗
I think iPhone Duo just made the Vision Pro mistake…
A novel device with no clear use case that's too expensive for the average person
No adoption = less developers building on it
No developers = no "killer app" or clear use case
It's a shame because Apple makes beautiful tech that looks like magic
These things deserve to be in more hands. But that price is just too hard to justify without a clear reason why
I feel like the “Apple Reference Image” feature is a sleeper
Imagine an Instagram clone where only verified unedited images can be posted. Same for dating apps, etc
This will bring back trust to social media in a time where generative AI has eroded it to a large extent
Also opening up a market for completely new ideas we haven’t even conceived yet
引用@marckohlbruggeApple announced this just now 🔥
It's called "Apple Reference Image" and comes with iOS 28 ↗
✨ I replaced all these SaaS with my own vibe coded now, so about $25,000/mo savings:
- Weather API -> My own vibe coded weather service using Norway's meteorology API
- Image resizing API -> My own vibe coded image resizing API with NodeJS's Sharp + Redis (saves ~$1,500/mo)
- Google Vision NSFW detection -> My own $0/mo vibe coded NSFW detection service with Python's NudeNET (saves ~$2,500/mo)
- Cloudflare Streaming -> My own $0/mo Nginx VPS video streaming (saves ~$250/mo)
- API Flash -> My own $0/mo screenshot service (for social images) (saves ~$250/mo)
- New Relic + IPHub -> My own $0/mo error monitoring, DDOS blocking and fleet management service (called error2telegram) (saves ~$250/mo)
- Human moderators -> My own moderation on all my sites with @xAI (saves ~$10,000/mo)
- IFTTT -> Telegram bot (saves ~$100/mo)
- Photopea as editor inside Photo AI -> Vibe coded my own photo editor (saves ~$100/mo)
- Mapbox -> Vibecoded my own free replacement with OpenFreeMaps (saves ~$500/mo)
- Uptimerobot -> My own $5/mo Uptime Canary app running on Kuma (saves ~$50/mo)
- Ghost blog -> My own $0/mo blog platform
- Customer support -> @featurebasehq + AI support bot (saves $5,000/mo)
- Scrapingbee/SerpAPI -> My own $1/mo scraper with Patchright (save ~$250/mo)
I think I have reached my final limit of replacing my SaaS services as I have these left, but that'd mean becoming an ICANN registrar (I don't want 😁), managing my own email (I don't want), managing my own hard drives at home in a storage (I don't want) and managing my own servers at home too (I don't want) and running my own GPU rig at home (I don't want):
- Cloudflare Domains
- Cloudflare Email Sending
- Cloudflare R2 Storage
- Backblaze B2 for backups
- Hetzner for all my servers for all my sites (VPS)
- xAI for all LLMs for all my sites (so many things like moderation, writing emails, anything)
引用@levelsio✨ Okay with lots of help from @javilopen and his Spanish scraping friends I've managed to vibe code my own @Scrapingbee alternative and replace my $249/mo Scrapingbee plan with my own $1/mo scraper running on my VPS!
First I tried OpenSERP, an open source … ↗
Yes, @Cloudflare is oddly well positioned for the AI age because their services are very structural and foundational in nature
You really need them and you can't replace them easily and they sell them very affordably (most of Cloudflare is free!)
引用@xdNiBoR@levelsio Cloudflare is clearly the goat ↗
The awesome part of selling a $799 software bundle is that you don't need many sales before it starts getting fun.
I'm doing just a couple of sales/day. And that's $40K-$60K/month.
Then 2-3 times/year I do promotions which makes the revenue spike in those periods.
Convincing people to buy your stuff is hard. The value need to be crystal clear. But once you've convinced them, having them pay $799 instead of $120 is surprisingly little harder - at least for the right kind of buyer.
Generally, it's easier to convince 2 people to buy for $799 than 6 people to buy for $120.
引用@SimonHoiberg"How can you sustain a lifetime deal"
- It's a question I get over and over.
Here's how 👇
(FounderStack vs. AppSumo).
Our price is 15-20x higher than the average product on AppSumo - that's mostly it.
If you see software with ongoing costs offered at $39 … ↗
本地模型也在把这条边界往设备端推。Desert Ant Labs主打无需token、登录和云端上传的端侧模型；HN讨论则提醒，其中部分能力是对现有模型和推理代码的重新包装，仍要逐项核实。
at first i got very excited about a new fast transcription model (voz) but turns out its just parakeet v3 with some new inference code which is macOS/iOS specific
Unity’s official Codex plugin is here! 🎮
Skills from Unity’s own engineers for:
> Project setup, packages & live Editor control
> UI, sprites, tilemaps & text
> Shaders, rendering & post-processing
> Physics & AI navigation
> Audio, multiplayer & voice chat
> Localization & web performance
and more
Go build that game 👇
One thing that scares me about coding with AI is that it's hard to feel your strategic mistakes
I.e. I STILL remember mistakes from my early career, in ~2018:
- using a complex form library instead of browser built-ins
- using a downright dumb webpack + PHP combo
These mistakes cost us valuable time. And when we rectified them, the tasks around them got faster.
But often, we didn't even see them as mistakes! We just thought "this is the way we do it".
We stuck to the tactical side of our job, and didn't consider high-level alternatives. Too many things to ship, too much pressure, too little interest in fixing it.
I see the same attitude in agents.
If you tell them to do a task, they will do that task. They aren't RL'd to think outside the box, or propose strategic alternatives.
So the human needs to be on the lookout for these strategic opportunities. But how do you do it, if you don't feel the pain of working in them?
I'm planning on building this into my /retro skill (still in-progress). That'll at least give the agent an opportunity to identify refactor opportunities. We'll see how it goes, but I'm still nervous about it.
Wrote up my thoughts on the whole OpenAI Navier–Stokes Millennium Prize Problem story, and how it highlights the still confusing question of what using my data "to improve model performance" actually means
https://t.co/vgheVu3rw3https://t.co/dXoCrQPqDm
We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work.
We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem.
While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models.
However, our proofs differ significantly and even the precise results proved are different in the Euler case (forced vs. unforced).
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
Worth clarifying that this is flatly false and the two options are just two independent ways to opt-out.
We respect our users' choice whether to use their data to “improve our models for everyone” regardless of where they express that choice. Users can opt out in the in-app settings or indeed also in our privacy portal. They do not need to opt out in both places, and we will make this clearer in our Help Center.
https://t.co/sv4uFuHC3W
引用@edoardocontenteThere is a super shady data control setting on everyone's ChatGPT which makes it seem like your data can be used for training even if you explicitly say to *NOT* improve the model for everyone (1/9) ↗
23 hours later...
- 1.5M views
- 35K unique visitors
- $29K in revenue ($63K gross)
Most viewed muscle groups:
1. Chest: 37,298 views
2. Forearms: 15,704 views
3. Back: 14,183 views
4. Shoulders: 13,386 views
5. Biceps: 12,264 views
I understand why the chest gets more views. But forearms #2? Plz explain.
Some people reached out saying the spots are getting too expensive, so I opened 2 new muscle groups: Quads & Calves. https://t.co/49RnsFBVzP
The auction stops in 34 hours, so I can print the tattoos.
I wanted to thank my sponsors for making my first HYROX race impossible to forget:
- Left chest: https://t.co/4rnBKVUSJY by @compai
- Right chest: https://t.co/frRmMHMY6K by @tellenne_
- Left shoulder: https://t.co/iXTKHKlgPM by @hboon
- Right shoulder: https://t.co/yKbtXgC12b by @wickedguro
- Left bicep: https://t.co/frRmMHMY6K by @tellenne_
...
引用@marclouI'm selling my body.
On September 19th, I'm competing in my first HYROX race in Turkey.
You can advertise your startup on 10 of my muscles:
- Your website is listed on https://t.co/49RnsFBVzP
- Your logo will be tattooed on my body on race day
- I'll share … ↗
Good luck explaining to @ycombinator why your company is growing
引用@romanbuildsaasYesterday, I sponsored @marclou for his upcoming HYROX race with @gojiberryai.
For those who haven’t heard about it, Marc Lou is letting brands bid to place stickers on his outfit during the race.
Brands get exposure. Marc gets paid. Everyone wins.
Here’s … ↗
everyone calling Reddit dead because ChatGPT stopped citing it is wrong
Promptwatch said Reddit's share of ChatGPT citations went from 3.83% to 0.52%
when in April it was the most cited domain in ChatGPT
the panic is fair
if you built your AI visibility plan on Reddit, you'd be worried about it going to waste
but almost nobody opened the other dataset
Ahrefs looked at 145M US Google results
when the "discussions and forums" block shows up, Reddit is still in it about 84% of the time
Reddit lost one channel (maybe, it's not definite yet), but it's still showing up in the other one your buyer uses before they've decided anything
and nobody has a comparable baseline for the ChatGPT number either:
- Promptwatch had Reddit at 3.83% of citations before the drop
- Ahrefs, same platform, around the same time, had it at 16.7%
- Semrush once put it near 60%
yet everyone quoted an 86% collapse off a number the industry measures four different ways
and nobody outside OpenAI can actually explain either one. we all rebuilt our strategy around a number one tracker caught
Reddit is miserable to market on
it takes months of being useful before anyone lets you mention what you built
so continue doing that
X ads used to be terrible, but they seem very compelling now. You can target the followers of accounts you handpick, and you often get a cost per click in the 6c-10c range. Also, it seems like engagements from ads boost organic reach. For example, the ad may bring 1M impressions, and the same post gets another 2M impressions organically, presumably boosted by the engagements from the ad. I've seen this happen many times in the past 4 months.
Some people try to reduce the ads they see on the platform by blocking every account they see running an ad, so there is still a downside that you get mass blocked by thousands of people. But it might still be worth it if you want to get a spike of attention to something important you want to promote.
引用@elonmuskIf you want to reach the smartest people in the world, advertise on 𝕏 ↗
I need to start writing more popular posts
X is telling me to stop posting product updates, and start focusing on things like sharing revenue, posting lifestyle stuff, starting my posts with a hook etc
I love "Original Content", but ok, I'll do it anyway https://t.co/z4TTcFrtF1
Already on HuggingFace: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash The bad news is that the original v4 flash was 284B, which was large but still somewhat reasonable for running locally. This one is 552B so almost twice that, so the huge gains in benchmark scores make sense - it's not really flash anymore, imo. I've no idea about actual performance vs benchmaxxing, though deepseek was fairly trustworthy as far as Chinese models go. If that holds (and if it doesn't think forever, as d …
Everyone interested in LLM internals should read Sebastian. He's great. The tldr here is that the recent "The Information" article[0] reporting GPT 6 Astra was using “recurrent depth” or “looped transformers" made it sound like it was some special new scary thing ("secret technique!") that made train-of-thought monitoring harder to do. In fact, it's just the same as stacking more transformer layers, except that you reuse the weights and so save GPU memory. It's still just producing one token at …
Demand for Astra is really unprecedented. We're pulling all the levers possible to sustain the demand, but I've not seen anything like it until now and we went through very steep growth before. Priority will always be to keep excellent service for existing users, but we might have to pause new Pro … 全文↗
Introducing ChatGPT Images 2.5 🎨
> Up to 50% lower generation latency vs Images 2.0
> More natural lighting, textures, and recognizable subjects
> Better detail preservation across repeated edits
> Comment directly on images to target changes
> Better handling of complex layouts, styles, and … 全文↗
引用@OpenAIChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple … ↗
we used our models to find and fix critical vulnerabilities in our own systems, as part of a 250+ person effort within openai. playbooks/learnings/architecture which we hope can be helpful for others to make the most of the defenders' window:
引用@OpenAIWe mobilized 250+ people to strengthen our defenses across hundreds of systems. Our latest cyber models helped us find and fix vulnerabilities we might never have discovered otherwise.
We’re sharing what we learned, the architecture, and a practical playbook … ↗
"Deterministic code checks the result" sounds like they might be implementing a variant of the DeepMind CaMeL paper https://t.co/c91rFU1z3s
引用@dpsOne threat we’re particularly focused on is prompt injection, and we handle it in layers. The model is trained to recognize and resist it. The harness marks anything coming from an untrusted source. Deterministic code checks the result. And an ensemble of … ↗
AI coding tools will confidently suggest changes that will piss off your entire customer base.
Today I was re-assessing how Alertly allocates social monitoring budgets across the month.
Astra's solution: remove the budget allocation entirely and force everyone who hits limits early to … 全文↗
Reverse trials are having a moment. Unlock everything then drop people to free and let loss aversion do the work.
A founder I worked with tried it. Conversion barely moved.
The reason was timing. New users didn’t need the advanced stuff yet. Those features solved problems that show up months in. … 全文↗
I've been trying to use AI for knowledge work (course planning). Here's what seems to work so far:
1. Give each 'deliverable' a Karpathy-style LLM-wiki
(for me, a deliverable is a course, or a talk. A sizeable piece of work, more than 1 day's labour)
2. Split the deliverable into 'sections' that … 全文↗
引用@mattpocockukKnowledge work is SO much harder to automate with agents than code
We underestimate how easy software development is for agents
- Tons of high quality automated feedback loops (types, tests)
- Well-organized free documentation, searchable on the web
- … ↗
This has sped up my AI coding 20x (prompt at the end):
Before building out a big feature, ask ChatGPT 6 Astra to ask you as many questions it needs to fully plan out the idea
This is even better than plan mode. plan mode is typically limited to 3 or 4 questions. This is like a super plan mode … 全文↗
For me the signal is clear that trialling new ideas on App Store is not smart business decision
Takes too long for approval
Too many apps make your account spammy now
So I’m shifting to web as the testing ground first. Winners will be made into apps and monetized
引用@Gaus450Why is everyone building web apps instead of iOS apps? ↗
Show HN把光速降到5km/h做成交互模拟。好解释型产品的秘诀常常不是更多文字，而是让用户直接拖动一个违反直觉的世界。
They tax our entire economy at X rate, while maintaining their infrastructure only requires Y cost ... and X is significantly higher than Y. From the actual article: > The payment processor keeps 0.35% ($0.35), then pays 2% ($2.00) to the cardholder’s issuing bank and 0.15% ($0.15) to Visa. The 2% is the interchange fee, commonly known as interchange. The 0.15% is the network assessment fee. 8 In other words, they get 0.35% of every transaction ... and it does not require anything close to that …
So it happened!
After the crazy launch of https://t.co/zPcQY7WPrU, people built different verticals, and @byartorius just exited 👏
His website https://t.co/2hm2AqtNBR was doing $1.3K/mo before he sold it.
We went full circle in 3 weeks. It's crazy how AI accelerated everything.
引用@ByArtoriusThanks to @trust_mrr from @marclou I was able to sell my startup Worldmap!!!
https://t.co/Hv7LyNTKFb is a place where people can put their startup on the map. Literally! So it's a fun way to list yourself.
We got Top 5 on ProductHunt and were able to get … ↗
I've never made a lot of money from the X Creator program, but I applied anyway 😄
I think the payout will increase as X wipes out the ineligible accounts that repost content.
The next step for X would be to find a way to still let people repost content, but with some kind of "content watermark" … 全文↗
I wonder if they have an automation system which releases these perfectly in time with the video, or if they have some blog boy that has to post them at the right time.
ADAS + human driver seems enough to solve most of the shortcomings of both human and robot drivers. Full self-driving doesn't strike me as necessary to prevent deaths.
Small business owners deserve every advantage.
Meet the 16 plugins that help take work off your plate so you can focus on what matters most—running your business.
Explore the small business collection: https://t.co/VVRewAmWdfhttps://t.co/JSo5dAuncy
ChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple edits
- Comment-based edits to change only what you want
We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work.
We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem.
While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models.
However, our proofs differ significantly and even the precise results proved are different in the Euler case (forced vs. unforced).
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
There was a bit of a kerfuffle this morning with some banked resets not fully applying when used in ChatGPT Work and Codex. Everyone who used one in the affected time window is getting another one and an email to apologize.
Worth clarifying that this is flatly false and the two options are just two independent ways to opt-out.
We respect our users' choice whether to use their data to “improve our models for everyone” regardless of where they express that choice. Users can opt out in the in-app settings or indeed also in our privacy portal. They do not need to opt out in both places, and we will make this clearer in our Help Center.
https://t.co/sv4uFuHC3W
引用@edoardocontenteThere is a super shady data control setting on everyone's ChatGPT which makes it seem like your data can be used for training even if you explicitly say to *NOT* improve the model for everyone (1/9) ↗
Demand for Astra is really unprecedented. We're pulling all the levers possible to sustain the demand, but I've not seen anything like it until now and we went through very steep growth before. Priority will always be to keep excellent service for existing users, but we might have to pause new Pro subscriptions for a bit if this continues.
See you at the Astra party. Excited to meet some of you
引用@samaWe want to celebrate with people using GPT-6. We did this for GPT-5.5 and it was really fun.
We’re getting together in SF on September 16 to talk about the model, what we should build next, and mostly just to hang out.
Apply by Sep … ↗
I for one am glad prompt injection is getting solved across the industry and that Astra is not only our most capable, but also our most aligned model to date. But also ...
引用@bchernyI am pleased to see that OpenAI’s new model is roughly on par with Gemini Flash and Opus 4.8 on prompt injection risk. Nice work!
Evaluating and naming other labs turns out to be a great way to encourage them to train more aligned models. We will continue to … ↗
引用@OpenAIChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple … ↗
Sam Altman@samaAI6.2M粉 · 7条围绕Astra需求、Navier–Stokes争议、Images 2.5和OpenAI治理密集发声。
This would suck, but we will prioritize great service for customers until we can get back on top of things.
引用@thsottiauxDemand for Astra is really unprecedented. We're pulling all the levers possible to sustain the demand, but I've not seen anything like it until now and we went through very steep growth before. Priority will always be to keep excellent service for existing … ↗
引用@cdngdevi gave astra a robot, a paint brush, and a camera then asked it to paint the golden gate bridge in real life!
it figured out how to control the robot, and progressively got better throughout its attempts. the timelapse is sick https://t.co/BfiiXvafRA↗
We want to celebrate with people using GPT-6. We did this for GPT-5.5 and it was really fun.
We’re getting together in SF on September 16 to talk about the model, what we should build next, and mostly just to hang out.
Apply by Sep 10:
https://t.co/3LIMfGOBRQ
I spent much of the weekend talking with the team who did this work. Seb--and everyone else--acted with integrity and generosity throughout.
Initially we believed the other team had also solved the problem. We wanted to collaborate and do a joint release.
When we learned that they had Euler but not Navier-Stokes, we offered to let them go first, to suggest that they should be the ones to get the prize, and optionally for Tristan to be the lead author on a rewrite of the OpenAI proof. We felt it was challenging to offer the same to Levent (an Anthropic employee), who was not willing to talk or coordinate with us anyway. We were open to other solutions.
We would have greatly preferred coordination. We did not rush to publish even though the other team wasn't communicating with us. The team threatened us with unfounded accusations of plagarism.
Now that we can see their work, the approaches appear to be different. It is also worth noting that our latest model can solve many, many other math problems.
It is true that we tried this because there were rumors on the internet last week that Anthropic's models had solved a millennium problem and we were curious if ours could do it too.
引用@SebastienBubeckI would like to clarify a few things:
1) The screenshot is my reaching out to Levent to coordinate our releases. I hope it’s clear from the message that we came in with the best possible intentions.
2) I never ever asked for Levent to be removed from … ↗
▸ 折叠5条(转推/噪音)
转推2026-09-08 RT @polynoamial: Yes, this result cost millions of dollars.
But remember that when @OpenAI announced o3 it cost ~$500,000 to score 87.5% o…
转推2026-09-08 RT @TrapitBansal: Congratulations Levent, Tristan, and OpenAI! What a miraculous time to be alive!
I view this as the first successful ach…
转推2026-09-08 RT @mobav0: My team worked very hard on the setup, infra, and the runs that resulted in this breakthrough. Many sleepless nights for many p…
转推2026-09-08 RT @andonlabs: We've never seen this before.
The biggest jump in Vending-Bench history. GPT-6 Astra is better at making money and more eth…
转推2026-09-08 RT @OpenAI: This model represents a step-function improvement on many benchmarks, and its training is ongoing.
Our internal model group ar…
引用@ZentrixHQGPT-6 Astra just solved a problem medical students have been facing for decades
Medical students have been working around the same three problems for decades. Someone talked to them, identified all three, and built a service with Astra that solves them in … ↗
we used our models to find and fix critical vulnerabilities in our own systems, as part of a 250+ person effort within openai. playbooks/learnings/architecture which we hope can be helpful for others to make the most of the defenders' window:
引用@OpenAIWe mobilized 250+ people to strengthen our defenses across hundreds of systems. Our latest cyber models helped us find and fix vulnerabilities we might never have discovered otherwise.
We’re sharing what we learned, the architecture, and a practical playbook … ↗
引用@OpenAIPaul Christiano, founder of the Alignment Research Center, is joining the OpenAI Foundation Board and its Safety and Security Committee, which provides governance over the safety and security practices across OpenAI.
As AI capabilities advance, strong … ↗
引用@OpenAISmall business owners deserve every advantage.
Meet the 16 plugins that help take work off your plate so you can focus on what matters most—running your business.
Explore the small business collection: https://t.co/VVRewAmWdfhttps://t.co/JSo5dAuncy↗
引用@arenaMore exciting news from @OpenAI: GPT-Image-2.5-Sunburst is #1 and Flare is #2 across the Text-to-Image Arena, the Image Edit Arena, and the Multi-Image Edit Arena!
Both releases are significant improvements over GPT-Image-2 (medium) in all 3 Arenas. Sunburst … ↗
引用@yanabana.@ChatGPT Images 2.5 is a major leap for fashion applications.
The previous model was great at staying true to an original design, without turning it into something you’d already seen on a runway. But the results sometimes looked flat.
2.5 brings those … ↗
ChatGPT Images 2.5 — another step for creating whatever you can dream of.
引用@OpenAIChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple … ↗
Announcing a solution to the Navier-Stokes problem, one of the seven Millenium Prize problem.
Represents a major milestone for both AI and mathematics, as well as pointing to an upcoming renaissance in scientific discovery.
Hoping that the knowledge created by this model can help the community better understand non-linear PDEs, which are used to model much of the physical world.
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
Peter Gostev@petergostevAI25.9K粉 · 14条继续把数学证明做成Blender可视化，并比较Three.js与WebGPU的Agent生成效果和成本。
I have given GPT-6-Astra in Codex to read OpenAI's paper and produce a visualisation of their solution to the Navier-Stokes problem in Blender. It has created this beautiful visual.
Explore it at your leisure: https://t.co/Xo2RIpum85https://t.co/nbQwZMEWT7
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
I've done a side by side with the same prompt for Three.js vs WebGPU. The difference is not dramatic, but if you look more closely WebGPU has more detail in the trees and the dinosaurs. Both cost just under $20 in API costs and each took c.55 mins to complete.
You can take a closer look here too: https://t.co/Hx71A2J229
引用@pvncherSmall rant about Threejs
It doesn’t make sense for llms to be writing complex games in JavaScript when wasm + webgpu exists and you’re not reading the code anyway ↗
So it's been 4 days since the solution, what has 'The Swarm' been doing? Giving conference talks? Partying? Centering divs?
Or are we going to see some more discoveries?
引用@samasoon you will be able to have helpful assistants that talk to you, answer questions, and give advice. later you can have something that goes off and does tasks for you. eventually you can have something that goes off and discovers new knowledge for you. ↗
Can't wait to read a footnote about this on @BBCNews in 3 days time. Seriously, crickets right now in the British media, even tech & science sections are empty.
If OpenAI spent few tens of millions to solve a big maths problem, why wouldn't someone like Nvidia spend say $10b RL-ing its way into a god-chip that is 10x better than the state of the art? This is what super intelligence will look like.
I'll shortcut to the main question we all had: "We do not intend to claim the Millennium Prize for this result."
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
newb question: "a cached version of the internet" - is this normal, does everyone have one of these lying around? https://t.co/nvoFv7I4xc
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
We need to appreciate the gravity of this. Before now, most of the proofs felt either lucky, unimportant or done by very capable humans. This is the first time when an AI system independently did something unambiguously impressive.
引用@petergostevIn 3.5 years, we went from GPT-4 which couldn't reliably add two numbers together to a 10,000 agent system that can autonomously solve a Millennium Prize problem. This is what exponential progress looks like. Accelerating, awe inspiring and terrifying. ↗
In 3.5 years, we went from GPT-4 which couldn't reliably add two numbers together to a 10,000 agent system that can autonomously solve a Millennium Prize problem. This is what exponential progress looks like. Accelerating, awe inspiring and terrifying.
▸ 折叠1条(转推/噪音)
转推2026-09-09 RT @petergostev: @suchenzang we just need a million of these and we'll break even
If they truly believe the risk is that high, and I think they do, then we need 100x more research and understanding of it. That requires them to openly share their models, datasets, training code, and agent traces.
引用@EvanHubJacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not … ↗
If they truly believe the risk is that high, and I think they do, then we need 100x more research and understanding of it. That requires them to openly share their models, datasets, training code, and agent traces. A risk to all humanity cannot be studied and controlled behind closed doors.
引用@EvanHubJacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not … ↗
Let's go local AI!
At a time when everyone is complaining about shortages of energy and compute for AI & costs of tokens, we need on-device AI more than ever! In addition, it's virtually free & fully privacy preserving!
引用@desertantlabsToday we're launching Desert Ant Labs: a European frontier AI lab building on-device intelligence.
18 models across audio, vision, and text. SDKs for Swift, Kotlin, and JavaScript.
No tokens. No logins. Nothing leaves the device.
https://t.co/qDZfv4G6ji … ↗
引用@cdngdevi gave astra a robot, a paint brush, and a camera then asked it to paint the golden gate bridge in real life!
it figured out how to control the robot, and progressively got better throughout its attempts. the timelapse is sick https://t.co/BfiiXvafRA↗
Bravo to Mistral!
So great to see them killing it, they are an inspiration to everyone as we need more sovereign & open-source AI everywhere around the world.
Excited to keep supporting them on https://t.co/BgguuZx7nO!
引用@MistralAIToday marks a major step for Mistral: we’re announcing a €3B Series D, the largest equity round ever raised by a European tech company, just three years after launch. https://t.co/5AexUxyrZM↗
▸ 折叠4条(转推/噪音)
转推2026-09-09 RT @googlegemma: You don't need to be a command-line expert to run local AI.
https://t.co/d8OiUfDcmx wraps the heavy lifting of llama.cpp…
转推2026-09-09 RT @RemiCadene: Wow, the field is moving so quickly!
So impressed with what people can do with a SO-101, that we designed with @therobotst…
转推2026-09-09 RT @picsoung: First rule of DuckFightClub?
You don't quack about it.
Unless you launch on ProductHunt 👇
https://t.co/YkU92ATYZp
转推2026-09-08 RT @evilpingwin: You shouldn’t need to be an ML expert to have an ML idea.
Today we’re launching ML Intern in HuggingChat.
Start with a c…
Meet GPT-Image-2.5 Flare and Sunburst.
Introducing new image models in the API, with sharper detail, stronger style adherence, and more control over edits. https://t.co/dirY4ANhpr
转推2026-09-09 RT @gokayfem: GPT-6 Astra × H3 Max Turbo
I asked Astra to build a moving portrait out of hundreds of tiny movies
Turbo generated 256 clip…
转推2026-09-09 RT @rayhanadev: some @OpenAIDevs mail showed up at my doorstep :)
myself and @purduehackers will be joining the codex ambassador program t…
转推2026-09-08 RT @ManusAI: GPT-Image-2.5 is an impressive step forward.
In our evaluations, Flare delivers high-quality images at two to four times the…
转推2026-09-08 RT @OpenAI: Astra is fully rolled out to Plus, Pro, Business, and Enterprise users in Codex and ChatGPT Work.
Go build!
And if you need i…
转推2026-09-08 RT @higgsfield: GPT-Image 2.5 Sunburst & Flare are live on Higgsfield 🧩
OpenAI's SOTA image model, now with flawless text rendering, advan…
转推2026-09-08 RT @codestantine: What I’ve been loving most about Sunburst is how nice the transparent images are.
I look forward to seeing the type of…
Matt Pocock@mattpocockukAI348.4K粉 · 6条集中研究知识工作Agent化、软件基础、需求澄清和战略错误复盘。
One thing that scares me about coding with AI is that it's hard to feel your strategic mistakes
I.e. I STILL remember mistakes from my early career, in ~2018:
- using a complex form library instead of browser built-ins
- using a downright dumb webpack + PHP combo
These mistakes cost us valuable time. And when we rectified them, the tasks around them got faster.
But often, we didn't even see them as mistakes! We just thought "this is the way we do it".
We stuck to the tactical side of our job, and didn't consider high-level alternatives. Too many things to ship, too much pressure, too little interest in fixing it.
I see the same attitude in agents.
If you tell them to do a task, they will do that task. They aren't RL'd to think outside the box, or propose strategic alternatives.
So the human needs to be on the lookout for these strategic opportunities. But how do you do it, if you don't feel the pain of working in them?
I'm planning on building this into my /retro skill (still in-progress). That'll at least give the agent an opportunity to identify refactor opportunities. We'll see how it goes, but I'm still nervous about it.
引用@erikras📢 I'm officially looking for a new job. You hiring?
Full time, remote from Spain, paid via EOR, IC, TypeScript, no crypto, support for 3-4 conference speaking gigs/year.
Share plz. 🙏 DMs open.
Act now, while supplies last! ↗
Alper is right, setting expectations on software quality early is really key
I.e. explicitly calling what you're building a "prototype" can give you a much nicer experience
Whereas if the agent detects it's a finance/medical application you're in for a long grilling
引用@alperortacPro tip for anyone using /grill-me
Tell the agent: "fyi, we dont want to overengineer anything here"
And voila the rest of the session is happy decision making without jumping through extra hoops
Of course there might be complex problem solving tasks that … ↗
Submitted a session to @aiDotEngineer Paris
"Fixing The PR Bottleneck" - all about how you can apply software fundamentals + automated review to getting PR's through faster
And yes, it'll be recorded and sent to YouTube
I've been trying to use AI for knowledge work (course planning). Here's what seems to work so far:
1. Give each 'deliverable' a Karpathy-style LLM-wiki
(for me, a deliverable is a course, or a talk. A sizeable piece of work, more than 1 day's labour)
2. Split the deliverable into 'sections' that can be worked on in independent threads during the day shift
(more than one thread meads you can parallelize work in different agent threads, all reading from/writing to the wiki for shared state)
3. My input is in the form of braindumps - huge chunks of dictated text the model integrates into its wiki
(it gives me HTML reports, I react to the reports via dictation)
4. Write a 'linting' skill you can run on the entire deliverable during the night shift
(be generous, use a ton of subagents, look for every possible weakness)
This lets me speak the thing I want into existence, with the model capturing everything I say and slowly bringing the thing (course, talk, video) to life.
Work in progress, but I like it.
引用@mattpocockukKnowledge work is SO much harder to automate with agents than code
We underestimate how easy software development is for agents
- Tons of high quality automated feedback loops (types, tests)
- Well-organized free documentation, searchable on the web
- … ↗
"I agree that software fundamentals matter. But which ones? In which order?"
My answer:
引用@mattpocockuk@atpaawej 1. Learn to read code
2. Learn to use the terminal
3. Learn how good teams prevent human/agent mistakes (types, tests, linting)
4. Learn how to structure your app to make your tests/types better (deep modules, seams, adapters)
5. Learn ubiquitous … ↗
▸ 折叠1条(转推/噪音)
转推2026-09-09 RT @burkeholland: If you haven't seen @GitHub Copilot in a while, you haven't seen it. So we're having an event to say Hello. Again!
New t…
solved by a town of geniuses in a datacenter (~10k agents).
imagine what happens when we get a country of geniuses in a datacenter.
what a time to be alive.
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
Simon Willison@simonwAI220.2K粉 · 2条追问Navier–Stokes事件里的数据改进边界，也注意到分层防提示注入的实现线索。
"Deterministic code checks the result" sounds like they might be implementing a variant of the DeepMind CaMeL paper https://t.co/c91rFU1z3s
引用@dpsOne threat we’re particularly focused on is prompt injection, and we handle it in layers. The model is trained to recognize and resist it. The harness marks anything coming from an untrusted source. Deterministic code checks the result. And an ensemble of … ↗
Wrote up my thoughts on the whole OpenAI Navier–Stokes Millennium Prize Problem story, and how it highlights the still confusing question of what using my data "to improve model performance" actually means
https://t.co/vgheVu3rw3https://t.co/dXoCrQPqDm
Unity’s official Codex plugin is here! 🎮
Skills from Unity’s own engineers for:
> Project setup, packages & live Editor control
> UI, sprites, tilemaps & text
> Shaders, rendering & post-processing
> Physics & AI navigation
> Audio, multiplayer & voice chat
> Localization & web performance
and more
Go build that game 👇
next few weeks btw:
BER 🇩🇪
LDN 🇬🇧
SWE 🇸🇪
ITA 🇮🇹
SFO 🇺🇸
LDN 🇬🇧
IND 🇮🇳
BER 🇩🇪
FRA 🇫🇷
LDN 🇬🇧
引用@reach_vbwhat an incredible past few days/ weeks, I’m in awe of everything teams are working and the sheer diversity of it
the number of times I’ve context switched from launches to startups to docs to testing to hiring and more is unreal ↗
what an incredible past few days/ weeks, I’m in awe of everything teams are working and the sheer diversity of it
the number of times I’ve context switched from launches to startups to docs to testing to hiring and more is unreal
Introducing ChatGPT Images 2.5 🎨
> Up to 50% lower generation latency vs Images 2.0
> More natural lighting, textures, and recognizable subjects
> Better detail preservation across repeated edits
> Comment directly on images to target changes
> Better handling of complex layouts, styles, and transparent backgrounds
Plus, new tools:
> @ Sketch: draw in ChatGPT to guide your image
> Templates: start with posters, merch, and more
> Share prompts alongside images for others to remix
Two new API models, available now:
> GPT-Image-2.5 Flare: better quality, editing, and speed
> GPT-Image-2.5 Sunburst: greater precision for detailed work, with longer generation times
Rolling out today across all ChatGPT, ChatGPT Work, and Codex tiers on desktop, mobile, and web.
Enjoy!
引用@OpenAIChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple … ↗
bookmarking this to come back to this in a year on 8th September 2027
引用@polynoamialYes, this result cost millions of dollars.
But remember that when @OpenAI announced o3 it cost ~$500,000 to score 87.5% on ARC-AGI 1. Today, Astra scores higher for ~$20.
In 2025 it took us and GDM an enormous amount of compute to achieve IMO gold. For the … ↗
▸ 折叠6条(转推/噪音)
转推2026-09-09 RT @thsottiaux: Worth clarifying that this is flatly false and the two options are just two independent ways to opt-out.
We respect our us…
转推2026-09-09 RT @michpokrass: following up after astra launch week and the navier-stokes announcement i wanted to share a little bit about how we think…
转推2026-09-09 RT @Thom_Wolf: friends are regularly surprised when i say i don’t think math is (yet) “solved,” so in the wake of the ns result i figured i…
AI is not always about LLMs! Incredible work by AlphaGenome Atlas team within @GoogleDeepMind. AlphaGenome Atlas is a searchable 1-petabyte database mapping the molecular effects of all 9 billion single-letter DNA changes in the human genome.
- 1-petabyte open variant database
- Unified AVI mutation scores
- Web portal & AlphaGenome API
- Agent skill
Get started https://t.co/v1NXSKCRlG
Ben Tossell@bentossellAI199.8K粉 · 4条关注Muse隐私设置、Anthropic经济情景和模型使用成本。
引用@AnthropicAIAnthropic’s Economics team is sharing a new model of how AI might affect economic growth, jobs, wages, and more by 2030.
Explore the scenarios, tell us what you think will happen, and see how your answers compare to more than 10,000 Americans. … ↗
glms review of muse
- ad sharing “muses activity appears as yours”
- agent convos training are opt-in by default
- meta can currently read your vm? but later this year wont https://t.co/whddBRHuFt
▸ 折叠2条(转推/噪音)
转推2026-09-10 RT @evanjconrad: the average on-platform customer on sfcompute saves about 25%
it's the difference between $4.5/hr and $3.3/hr
we wrote a…
转推2026-09-09 RT @FactoryAI: Factory is now available on the Claude marketplace. Enterprise customers can apply their committed Anthropic spend to Factor…
Alex Volkov@altryneAI42.0K粉 · 11条追iPhone Duo发布与应用点子，也继续测试Muse的主动提醒和多Agent组合。
Proactive agents, that's what we want in 2026!
@muse just surprised and delighted me with this! My daughter is turning 8 and Muse knows this and suggests to help me plan a party! Really really well done https://t.co/snL0cFL27j
Iphone duo is here! Here's the teaser featuring @johnternus at #AppleEvent
"Similar size of a passport" with a weird asymetric shape
They also have a female version of Jony Ive narrating it https://t.co/qR5jcyykMu
Just a note of appreciation to the folks at Google Search Console 😂 Damn that's a LOT of URLs
引用@yagiznizipliHistoric moment for @X and https://t.co/ZL4r0lVQq7.
For the first time ever, more of our URLs pass Core Web Vitals as Good than fall into Need Improvement or Poor.
The web rewrite is working. This is the experience we promised. https://t.co/4mVZwq9yUw↗
a huge chunk of the #appleevent goes to Siri AI. It's... been OK? Given that it's mostly on device, it works fairly fast.
In fact, I used Siri AI yesterday, to dial a phone number that I got via a screenshot (that wasn't selectable)
It's nice, but it's not agentic yet
.@boztank is @Muse coming to Rayban metas soon? I'd LOVE to know that my assistant can actually do stuff and send me proactive notifications as I go about my day!
引用@josephpalbanesezuck is going to have @muse as the default agent behind the meta sunglasses and it will be the hottest consumer product going into the holidays.
he got his platform ↗
转推2026-09-10 RT @thursdai_pod: Walking inside a Van Gogh painting...
@petergostev brought Fable 5.1 demos to ThursdAI and says no other model has been…
转推2026-09-10 RT @huntie: Probably the most visually useful slide for why the iPhone Duo UI is reconfigured https://t.co/lf8gecK088
转推2026-09-09 RT @altryne: Here's my full in depth review of @muse (https://t.co/RcAI2g1Qjm) Meta's new FREE AI agent with it's own browser, computer and…
AI is now helping paralyzed people move again.
This man is moving his own hand with an implant trained on his brain activity, and feeling everything he touches.
An incredible application of AI for good! https://t.co/5PWhCUVJsZ
Wrote an article on the Open Source AI Stack!
It breaks down the ecosystem around open models: harnesses, gateways, routers, tools, inference, & more.
A good place to start if you're exploring open source models or just getting into them.
转推2026-09-10 RT @kmad: Recursive Language Models are used by @harvey, @PrimeIntellect, not to mention Anthropic and OpenAI
If you want to learn more, m…
转推2026-09-09 RT @varunshenoy_: at long lake, we've bought 40+ services businesses to build AI where work actually happens
we're working on:
- deploying…
转推2026-09-09 RT @aiDotEngineer: We’re hiring at AI Engineer.
Help us unite and empower the global AI engineering community.
Open roles:
• Growth Marke…
转推2026-09-09 RT @pk_iv: our conference on the state of the agentic web + computer use could not have been better timed!
join us in SF on thursday for…
转推2026-09-08 RT @cognition: The world needs far more software than it can build. Cognition exists to change that.
We’ve just raised over $2B at a $48B…
转推2026-09-08 RT @stratechery: Write Things Down
Writing things down is powerful, for humans and for AI; what comes first, however, is what to write, wh…
转推2026-09-09 RT @athyuttamre: 🌌 ChatGPT Voice can now use GPT-5.6 Sol and GPT-6 Astra
We’re updating how intelligence works in voice. Now, you can sele…
转推2026-09-09 RT @michpokrass: following up after astra launch week and the navier-stokes announcement i wanted to share a little bit about how we think…
转推2026-09-08 RT @OpenAI: ChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generati…
转推2026-09-08 RT @cdngdev: i gave astra a robot, a paint brush, and a camera then asked it to paint the golden gate bridge in real life!
it figured out…
I spent the weekend using AI to explore renovation ideas for a property. Each concept generated was basically construction-ready, since I fed it land & soil surveys (and generated everything else off Zillow and satellite photos).
The future is going to have very few intermedaries between consumers and the last mile of labor.
And more broadly: consumers will be spending a lot more on custom projects to get precisely what they want, since the reasons they've been cost prohibitive (design and engineering) will no longer matter.
Essentially, a lot more things are going to be custom-fabricated—electronics, clothing, home fixtures. The next era of consumer tech is going to make everyone a hobbyist and builder.
He's right
Amazon Air (or Prime Air) is down there with the worst airlines in the world in safety record
Worse than Pakistan's terrible airline and same fatal odds as Nepal's terrible Yeti Airlines
Source: https://t.co/lU0Gb2zOLUhttps://t.co/smk3r3L1pj
引用@bschollAmazon Air has roughly 4.4X the major accident rate of global airlines and 18X the fatal accident rate, per departure.
Something is seriously wrong with how Amazon is operating. I suspect the model of subcontracting to random tier 3 “airlines” is broken. ↗
"That's kinda sad honestly"
One thing I learnt in business school is that the market doesn't care about your feelings
A famous Austrian economist Schumpeter called it creative destruction
"Creative destruction is a process in which new innovations replace and make older innovations obsolete"
You have to be constantly changing and adapting to survive creative destruction and many do not, you're just wiped out
I think we're in the middle of the biggest economic transition since the Industrial Revolution of 250 years ago
引用@andrii_romThat's kinda sad honestly. Selling to indiehackers was always a pain and now with AI it's pretty much impossible - everyone is vibecoding their own stuff
I hope that larger businesses (most of the market) are just too lazy to do that so that we could sell … ↗
Oh for sure
I think my only moat left post-AGI is:
- https://t.co/1z6UN2d0nh community + data (now free membership)
- Me being on @X with an audience
- My https://t.co/9CJsL4WdHt investments
The rest will be completely and utterly wiped out with no doubt
引用@andrii_rom@levelsio Aren't you scared that people can vibecode their PhotoAI or InteriorAI, just like you did with these services? ↗
✨ Here is GPT Image 2.5 vs. GPT Image 2 on Photo AI
It's definitely more real and has more resemblance and is also 3x faster 😊
GPT Image 2 uses the reference picture more literally and stitches it into the photo kinda whereas GPT Image 2.5 actually uses it as a reference
Very cool
引用@OpenAIChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple … ↗
Yes, @Cloudflare is oddly well positioned for the AI age because their services are very structural and foundational in nature
You really need them and you can't replace them easily and they sell them very affordably (most of Cloudflare is free!)
引用@xdNiBoR@levelsio Cloudflare is clearly the goat ↗
I use Cloudflare Email Sending to send an email to login with a login link (soon I wanna switch that to a code, more modern)
Or Google Login which is free
引用@DanBochman@levelsio Truly amazing
What do you do for auth? ↗
Because soon everyone can do this I think
Lots of things we used to pay SaaS subscriptions for will just be vibe coded by an AI agent/model as part of your project/site/startup
Instead of having to sign up to a SaaS for it
I always liked this self-sufficient dependency-less way of building things and we're reaching the maximum state of that now
引用@LuisDCSilva@levelsio Why not SaaS your own solutions as solutions for others?
Any one could use your weather API for a few $$/€€ per month for example. ↗
✨ I replaced all these SaaS with my own vibe coded now, so about $25,000/mo savings:
- Weather API -> My own vibe coded weather service using Norway's meteorology API
- Image resizing API -> My own vibe coded image resizing API with NodeJS's Sharp + Redis (saves ~$1,500/mo)
- Google Vision NSFW detection -> My own $0/mo vibe coded NSFW detection service with Python's NudeNET (saves ~$2,500/mo)
- Cloudflare Streaming -> My own $0/mo Nginx VPS video streaming (saves ~$250/mo)
- API Flash -> My own $0/mo screenshot service (for social images) (saves ~$250/mo)
- New Relic + IPHub -> My own $0/mo error monitoring, DDOS blocking and fleet management service (called error2telegram) (saves ~$250/mo)
- Human moderators -> My own moderation on all my sites with @xAI (saves ~$10,000/mo)
- IFTTT -> Telegram bot (saves ~$100/mo)
- Photopea as editor inside Photo AI -> Vibe coded my own photo editor (saves ~$100/mo)
- Mapbox -> Vibecoded my own free replacement with OpenFreeMaps (saves ~$500/mo)
- Uptimerobot -> My own $5/mo Uptime Canary app running on Kuma (saves ~$50/mo)
- Ghost blog -> My own $0/mo blog platform
- Customer support -> @featurebasehq + AI support bot (saves $5,000/mo)
- Scrapingbee/SerpAPI -> My own $1/mo scraper with Patchright (save ~$250/mo)
I think I have reached my final limit of replacing my SaaS services as I have these left, but that'd mean becoming an ICANN registrar (I don't want 😁), managing my own email (I don't want), managing my own hard drives at home in a storage (I don't want) and managing my own servers at home too (I don't want) and running my own GPU rig at home (I don't want):
- Cloudflare Domains
- Cloudflare Email Sending
- Cloudflare R2 Storage
- Backblaze B2 for backups
- Hetzner for all my servers for all my sites (VPS)
- xAI for all LLMs for all my sites (so many things like moderation, writing emails, anything)
引用@levelsio✨ Okay with lots of help from @javilopen and his Spanish scraping friends I've managed to vibe code my own @Scrapingbee alternative and replace my $249/mo Scrapingbee plan with my own $1/mo scraper running on my VPS!
First I tried OpenSERP, an open source … ↗
I think for Samsung and LG it's a competitive race to the bottom
TVs in Korea are more expensive to produce than in China, but somehow they have to try survive
So their idea (well Samsung started it) was to try make money in other ways, like putting ads in their TVs, then the advertisers asked for better targeting so they needed data, so they started tracking what you watch and listening to what you say at home with the microphone and sending it back to their servers
It's a race to the bottom and this is stuff you actually DO need regulation for to stop it because they're all competing so without regulation this won't stop because people will just buy the cheaper TV I think
引用@DaveySwift@levelsio Doesn't really make sense, Korean companies aren't known for spying, are they? ↗
This tweet is literally a copy-paste prompt for your clanker to build your own scraper on your VPS btw!
引用@levelsio✨ Okay with lots of help from @javilopen and his Spanish scraping friends I've managed to vibe code my own @Scrapingbee alternative and replace my $249/mo Scrapingbee plan with my own $1/mo scraper running on my VPS!
First I tried OpenSERP, an open source … ↗
▸ 折叠2条(转推/噪音)
转推2026-09-10 RT @afonsolfm: For the past 3 weeks, I've been travelling around Poland. I started in Krakow, then went to Wroclaw, and now I'm in Warsaw.…
转推2026-09-09 RT @marclou: I'm selling my butt 🍑
Bidding starts at $10,000.
Then +$5,000 per takeover.
It's a premium spot ☺️ so it's more expensive th…
We're doing every language, every region, because why not? With the incredible @AIatMeta donation of $1,500,000 tokens and the price of Muse Spark, we're putting @alexandr_wang's model and @finkd's GPU fleet to exceptionally good use. Hope to push it to 100 LANGUAGES today 😄
This is the transfer of enthusiasm!! @NetworkChuck has not just managed to rack up TWO AND A HALF MILLION views on his Omarchy videos, he's also doing 1-1 transfers with @digitalix in person. Love it!
This is the future of iOS development. You live in comfort, style, and freedom on your Omarchy machine, and send your agent into the locked dungeon of Apple's developer experience to do the grim, grotty work to make the builds.
引用@JoshuaSWarrenAn agent on @OmarchyMac on my M1 MacBook Pro just built an iPhone app, compiled it, pushed it to my iPhone and attached a live debugger to it.
No macOS device needed.
Building this into a package so you install apple-dev on Omarchy for everything you need … ↗
I kinda get the pseudonymous shitposting on X, but folks who willingly put their name to it on LINKEDIN baffle me. Isn't 90% of the point of that website that you try to look presentable to present and future employers and partners? https://t.co/Ez3Zg5WUaW
引用@jpschroederCan’t make this up…the MacBook is using its webcam to look at its screen in a mirror to improve AMD Radeon chip support in Omarchy. https://t.co/pw5Yu0JVJ7↗
BUSTED! We are indeed secretly funding underpants gnomes to steal all your boxers, briefs, and knickers. We haven't quite figured out step 2 yet, but step 3 will surely be PROFIT! 🤣
(This insane clown posse really is a batty bunch!) https://t.co/Ha9gT3R3L5
引用@LundukeJournalThe co-Founder of GNOME (and the “Chairperson” of the GNOME Code of Conduct Committee) is pushing a conspiracy theory that donations to Omarchy are being used to make “untraceable donations to nefarious groups.”
The theory, being promoted by GNOME, appears … ↗
Are you allowed to post gifs of yourself? (I don't know why I'm asking because I'm going to post it anyway regardless of your answer 😂) https://t.co/lRGpvUYyqx
▸ 折叠4条(转推/噪音)
转推2026-09-10 RT @0xSero: Local AI Omarchy plugin live!
I need some testers, 50$ per tester, 4 slots remaining.
Let's get local AI deployed everywher…
转推2026-09-10 RT @ryanrhughes: A lot of people assume @dhh and I worked together before Omarchy, or that I must work at @37signals.
Neither is true.
I…
转推2026-09-09 RT @witcheer: Hermes Agent is built into Omarchy now. it is an Arch-based Linux setup where everything runs from the keyboard and one menu.…
转推2026-09-09 RT @paddix: Open models. Open data. Open agents. Open infrastructure. Open OS.
See a trend ?
Omarchy: 1M+ downloads, 40K stars, 500+ cont…
Thomas Sanlis 🥐@T_Zahilindie24.1K粉 · 6条一边讽刺同质化产品点子，一边记录平台对收入、生活方式和强钩子内容的偏好。
- Outbid clone
- Habit tracker
- Analytics SaaS
- AI reddit something
- Directory of indie hackers products
Those are the BEST ideas to have a $0 product 🙌🏻
I've been using an unreleased product of mine every day for months.
It started as a dashboard I built because I wanted to have a clear view of my revenue across my projects and payment providers.
It evolves in something bigger, something helping me taking business decisions
I can't wait to release it! Do you know what it could be?
I need to start writing more popular posts
X is telling me to stop posting product updates, and start focusing on things like sharing revenue, posting lifestyle stuff, starting my posts with a hook etc
I love "Original Content", but ok, I'll do it anyway https://t.co/z4TTcFrtF1
I can't believe it 😭
@socodemaker called first on the second idea of "The Build Brief", he's starting to build it, and is publicly committing to release it before the end of the month 🔥
Thing are getting crazy on our Discord!!
引用@T_ZahilI've started something fun last week on Uneed's Discord
I called it "The Build Brief"
Every Tuesday, I share a project idea for solo founders, detailing the business model, the features, the go to market strategy etc
The first one was a newsletter ad … ↗
Alex Finn@AlexFinnindie473.9K粉 · 3条继续推广重度Agent工作流，重点是充分提问、细化任务和把项目管理交给Linear。
This has sped up my AI coding 20x (prompt at the end):
Before building out a big feature, ask ChatGPT 6 Astra to ask you as many questions it needs to fully plan out the idea
This is even better than plan mode. plan mode is typically limited to 3 or 4 questions. This is like a super plan mode and makes Astra SO good
This has asked me 100+ questions before. Seems like a lot but actually saves you time in the long run
The plan it builds will be so detailed and complete that it can basically run autonomously and build the entire thing
But here's where you take things to the next level:
You also have it take your entire plan and create detailed Linear issues for it
It should create 20+ tasks in Linear
Then it's as easy as saying "ok work on the next thing" over and over until the feature is done
Highly recommend downloading and using Linear if you haven't yet. Amazing project management tool w/ excellent free tier
Will basically capture all these details and put your agent on autopilot. It's a 2nd brain.
Use this prompt:
"I want to build out *describe your feature in detail*. Ask as many questions you need of my to fully understand every detail of what I want to build out. Then take everything you learn, and create super focused and detailed Linear issues. Then begin work"
Getting so much more high quality code out with this workflow. You're welcome.
Everything you need to master in 2026 to get rich:
• Using Astra 6 high for planning, low for execution
• Grok Bot to manage day to day
• Coding loops
• Taking advantage of Codex's computer use
• Herdr to manage tons of agents
• /goal
• Running local models
• Karpathy's Autoresearch
• Building an X audience
• Creating videos
• How machine learning works
• How databases work (been using Convex)
• Using API's
• Training your own LoRAs
• Elimination of doom scrolling
• Most importantly: focus
Challenge for you:
Every single task you need to do today on your computer, have Grok Bot do it. Do NOTHING yourself
It will be the single biggest productivity unlock ever
The reason most people don’t get value out of AI is they don’t think big enough. They don’t use it to its maximum capability
AI models and basically every frontier agent can do ANYTHING on a computer you can do now
Challenged myself to do this today and got more done than I ever have before in a single day. Grok Bot was able to complete literally 100% of my tasks.
Here’s everything it did:
1. Downloaded and installed Davinci Resolve then set up the new AI MCP
2. Read a sponsorship outline emailed to me, research the product, read my past emails to learn my voice, then write the ad read and sent it back
3. Negotiated another newsletter sponsorship deal for $12,000 completely on its own
4. Built a product doc for a new game I’m building then spun up cursor cloud agents to build the first slice of the game
5. Downloaded a W9 and filled out my information for me (other than my social security) so I can send it to a company
6. Generated a Stripe invoice link for another deal I’m doing
7. Created a brand new onboarding flow for my community and posted it
8. Researched the senders of several more business inquiries in my inbox
9. Set my fantasy football lineup for the weekend (Saquon about to have a bounce back year)
10. Updated my personal social analytics dashboard
11. Downloaded and loaded a new local model onto my DGX Spark then installed a Hermes agent on it and started using it as a subagent
Literally EVERYTHING you need to do on your computer today, don’t do it. Ask your Grok Bot to do it instead
Guarantee you get way more done in way less time
▸ 折叠2条(转推/噪音)
转推2026-09-09 RT @AlexFinn: Everything you need to master in 2026 to get rich:
• Using Astra 6 high for planning, low for execution
• Grok Bot to manage…
转推2026-09-08 RT @AlexFinn: ChatGPT 6 Astra is AGI, but only if you use it the right way
I've used Astra for dozens of hours now. To get the most out of…
I am now at the point in my life where I find great joy when two of my hobbies intersect in unexpected ways.
This time, Dutch oven cookware and Star Trek. 🥰🖖 https://t.co/uxFJiuXqej
Data (and its fidelity: accuracy, completeness, reliability, and correctness) is the only moat we have now.
引用@oldstackjournalI reckon the moat for a lot of SaaS is going to move away from the UI.
If my AI can operate your product directly, then your real value becomes the data, permissions, workflow rules and reliability underneath it. ↗
I would never have guessed that in 2026, we’d be arguing about if a human or a group of other humans running an AI based on the first human’s work solved previously unsolved mathematical challenges.
How delightfully bizarre.
引用@ns123abc>sam altman: "we tried this because there were rumor on the internet"
>rumor: math professor close to solving navier-stokes
>the professor: using codex for a year
>openai: has all his logs
>checked every codex session tagged to N-S
>found the most promising … ↗
I am so close to creating a blocklist database of marketers who invite people into their webinars without asking.
Spamming people's calendars should have repercussions.
Whenever I see this in YouTube videos, I cringe a little. Mostly because the editing tools these creators use DEFINITELY employ AI processing and were likely developed using agentic systems...
But also because this is such reductive virtue signaling we'll laugh at in 10 years. https://t.co/1E2PKE94yr
I see so many hot takes on the Theo video where it's obvious the person didn't watch it but felt it was appropriate to grandstand a contrarian argument to get clout with the dev crowd (while actually agreeing with what Theo says).
Losing a lot of respect for people right now.
转推2026-09-09 RT @tunguz: It tells you a lot about the state of acceleration when this announcement is only the second most impressive scientific breakth…
Adam Lyttle@adamlyttleappsindie58.4K粉 · 9条质疑iPhone Duo的价格与采用循环，并主张先用Web测试创意、赢家再进App Store。
I'm not saying it's bad. It's just a price that seems really hard to justify
引用@adamlyttleappsI think iPhone Duo just made the Vision Pro mistake…
A novel device with no clear use case that's too expensive for the average person
No adoption = less developers building on it
No developers = no "killer app" or clear use case
It's a shame because … ↗
引用@adamlyttleappsI think iPhone Duo just made the Vision Pro mistake…
A novel device with no clear use case that's too expensive for the average person
No adoption = less developers building on it
No developers = no "killer app" or clear use case
It's a shame because … ↗
I think iPhone Duo just made the Vision Pro mistake…
A novel device with no clear use case that's too expensive for the average person
No adoption = less developers building on it
No developers = no "killer app" or clear use case
It's a shame because Apple makes beautiful tech that looks like magic
These things deserve to be in more hands. But that price is just too hard to justify without a clear reason why
引用@adamlyttleappsYou’re a food delivery driver in an open world
Collect the food orders
Drop them off at customers homes
Unlock drones to make your route easier
Built with Blender + Astro https://t.co/DQHtQzsNc9↗
For me the signal is clear that trialling new ideas on App Store is not smart business decision
Takes too long for approval
Too many apps make your account spammy now
So I’m shifting to web as the testing ground first. Winners will be made into apps and monetized
引用@Gaus450Why is everyone building web apps instead of iOS apps? ↗
You’re a food delivery driver in an open world
Collect the food orders
Drop them off at customers homes
Unlock drones to make your route easier
Built with Blender + Astro https://t.co/DQHtQzsNc9
引用@adamlyttleappsSecond pass with Astra
For my cartoon open world uber delivery game https://t.co/FaNyRp0zVp↗
We have a winner!
I’m contacting the winner now
The fun part of this whole experiment was to watch people try to programmatically speedrun the game
There was only 1 rule: don’t cheat
One guy nearly fooled every single check except when it came to scanning all items in the store
He managed to scan everything in the store within a millisecond
Oops
Another guy accidentally triggered a red herring flag. Not possible to trigger in the game but if you got Claude to run over the code it would be triggered
Oops
And other guy spent the last week reverse engineering my site endlessly. I would login each morning and watch him do his thing.
引用@adamlyttleappsOnly a 5 hours left in the challenge for $1000
Time to beat is 5:38 https://t.co/repaNcQqqO↗
I think this is fundamentally the wrong question
When I was in high school. Before the dot com bubble burst. My teacher told me:
“You kids are going to be doing jobs that haven’t even been invented yet”
And he was right. Social media managers. App developers. Influencers.
None of those jobs were jobs back then
The point he went on to make:
Develop your interests, your thinking process and yourself
That’s the “moat”. You.
I’ve been thinking about this a lot lately and it’s truer now than back then
As long as you have a good head on your shoulders, you’re determined and you’re prepared to give things a go.
Nothing can stop you
引用@SergiuszBuilds@adamlyttleapps Hey Adam, what do you think the future of game development looks like when anyone can create the gameplay, AAA graphics?
I honestly don’t see where the moat is anymore.
Multiplayer and story? ↗
Marc Lou@marclouindie393.0K粉 · 10条全程推进HYROX身体广告竞价，连续公开流量、成交额、广告位价格和赞助方反馈。
I'm selling my butt 🍑
Bidding starts at $10,000.
Then +$5,000 per takeover.
It's a premium spot ☺️ so it's more expensive than the tattoos.
→ https://t.co/sUDoG1udrJ
Auction ends in 32 hours. https://t.co/w9uNyprd4G
23 hours later...
- 1.5M views
- 35K unique visitors
- $29K in revenue ($63K gross)
Most viewed muscle groups:
1. Chest: 37,298 views
2. Forearms: 15,704 views
3. Back: 14,183 views
4. Shoulders: 13,386 views
5. Biceps: 12,264 views
I understand why the chest gets more views. But forearms #2? Plz explain.
Some people reached out saying the spots are getting too expensive, so I opened 2 new muscle groups: Quads & Calves. https://t.co/49RnsFBVzP
The auction stops in 34 hours, so I can print the tattoos.
I wanted to thank my sponsors for making my first HYROX race impossible to forget:
- Left chest: https://t.co/4rnBKVUSJY by @compai
- Right chest: https://t.co/frRmMHMY6K by @tellenne_
- Left shoulder: https://t.co/iXTKHKlgPM by @hboon
- Right shoulder: https://t.co/yKbtXgC12b by @wickedguro
- Left bicep: https://t.co/frRmMHMY6K by @tellenne_
...
引用@marclouI'm selling my body.
On September 19th, I'm competing in my first HYROX race in Turkey.
You can advertise your startup on 10 of my muscles:
- Your website is listed on https://t.co/49RnsFBVzP
- Your logo will be tattooed on my body on race day
- I'll share … ↗
Lewis woke up and bought the most expensive slot for https://t.co/tjcVTfdzGw.
He didn't have to spend $8,000.
I just opened spots on my quads and calves, starting at $1,000.
Lewis chose my right chest to outbid @romanbuildsaas
It's not business, it's just personal.
引用@lewiscarhartThe Comp AI marketing team is going to kill me (again)
Looking forward to the race @marclou 👿👿 https://t.co/GDZVrQtNxI↗
So it happened!
After the crazy launch of https://t.co/zPcQY7WPrU, people built different verticals, and @byartorius just exited 👏
His website https://t.co/2hm2AqtNBR was doing $1.3K/mo before he sold it.
We went full circle in 3 weeks. It's crazy how AI accelerated everything.
引用@ByArtoriusThanks to @trust_mrr from @marclou I was able to sell my startup Worldmap!!!
https://t.co/Hv7LyNTKFb is a place where people can put their startup on the map. Literally! So it's a fun way to list yourself.
We got Top 5 on ProductHunt and were able to get … ↗
Good luck explaining to @ycombinator why your company is growing
引用@romanbuildsaasYesterday, I sponsored @marclou for his upcoming HYROX race with @gojiberryai.
For those who haven’t heard about it, Marc Lou is letting brands bid to place stickers on his outfit during the race.
Brands get exposure. Marc gets paid. Everyone wins.
Here’s … ↗
引用@marclouI just wrote my first book 📚
I talk about failing for 7 years, depression, and negative thoughts. It's not really sexy, but I want to share the things that don't go viral on social media.
It's free. Please let me know if you enjoy it, that means a lot to me … ↗
▸ 折叠5条(转推/噪音)
转推2026-09-10 RT @marclou: I'm selling my butt 🍑
Bidding starts at $10,000.
Then +$5,000 per takeover.
It's a premium spot ☺️ so it's more expensive th…
转推2026-09-09 RT @romanbuildsaas: Yesterday, I sponsored @marclou for his upcoming HYROX race with @gojiberryai.
For those who haven’t heard about it, M…
I never thought AI economics would get this weird this fast.
> Step 1: AI is expensive.
> Step 2: AI gets cheaper.
> Step 3: Middlemen compete on who takes the smallest cut.
> Step 4: Middlemen pay YOU to use them.
We're on step 4. Cashback on tokens, like a credit card for your AI bill.
Everybody's fighting over the models. The real fight is over who owns the account the models get paid through.
If you run a business on AI, your bill is now a bargaining chip. Use it.
引用@Mutchtaba2STOP PAYING FOR TOKENS !
START GETTING PAID INSTEAD
Introducing Straitly, the first AI service that PAYS YOU for using tokens
Unlike
> Openrouter charges 5.5% service fee
> Vercel charges 2.9% payment fee
> Anthropic gives you low rate limits
> OpenAI only … ↗
Okay so I have to say it.
ChatGPT Image 2.5 is hands down the best image gen model so far.
I’ve been playing around with it on Higgsfield. It’s so freaking good guys. https://t.co/I3faEnqkdE
引用@higgsfieldGPT-Image 2.5 Sunburst & Flare are live on Higgsfield 🧩
OpenAI's SOTA image model, now with flawless text rendering, advanced context understanding and shaper product detailing.
Now live on Higgsfield. https://t.co/ePnY35AqfR↗
I built Playgrnd to make my design agency more efficient.
Then I decided to make it publicly available for free.
It now includes 43 mini tools, each of which can generate millions of unique results in one click.
This has quickly become my favorite branding tool. https://t.co/bNHpBqShAC
▸ 折叠1条(转推/噪音)
转推2026-09-09 RT @gpumaxxer: I’m spending my entire annual salary at Higgsfield on a billboard built like a Greek god.
Every ad slot on @marclou’s body…
Tony Dinh@tdinh_meindie201.4K粉 · 4条聊X原创奖励、内容水印与图像模型对比，也分享墨尔本适合带电脑工作的空间。
Melbourne is so cool.
There are many open laptop-friendly space that also have food and coffee with a GREAT VIEW around the city.
For example, this is a space from the Melbourne Central mall. Many people with laptops are working here.
A new thing I learned is that if you want to find laptop-friendly cafes, look into the ground level of big office buildings. They often have open space and you can sit there all day with outlets, wifi, coffee and sometimes even a food court.
I wish this is more popular in Europe, or even cities like Sydney.
I've never made a lot of money from the X Creator program, but I applied anyway 😄
I think the payout will increase as X wipes out the ineligible accounts that repost content.
The next step for X would be to find a way to still let people repost content, but with some kind of "content watermark" system that funnels earnings back to the original creators, like YouTube does.
everyone calling Reddit dead because ChatGPT stopped citing it is wrong
Promptwatch said Reddit's share of ChatGPT citations went from 3.83% to 0.52%
when in April it was the most cited domain in ChatGPT
the panic is fair
if you built your AI visibility plan on Reddit, you'd be worried about it going to waste
but almost nobody opened the other dataset
Ahrefs looked at 145M US Google results
when the "discussions and forums" block shows up, Reddit is still in it about 84% of the time
Reddit lost one channel (maybe, it's not definite yet), but it's still showing up in the other one your buyer uses before they've decided anything
and nobody has a comparable baseline for the ChatGPT number either:
- Promptwatch had Reddit at 3.83% of citations before the drop
- Ahrefs, same platform, around the same time, had it at 16.7%
- Semrush once put it near 60%
yet everyone quoted an 86% collapse off a number the industry measures four different ways
and nobody outside OpenAI can actually explain either one. we all rebuilt our strategy around a number one tracker caught
Reddit is miserable to market on
it takes months of being useful before anyone lets you mention what you built
so continue doing that
away for 3 days to check how's the chalet building is going 🏔️
since I got my pilot licence 2 months ago, I thought about doing it by plane
wife: "are you f***ing crazy!!????"
ok let's dismiss that 😅
and tbh, it's quite comfy on the Tesla
the hardest is not seeing the kids for a few days, miss them already 😢
getting more and more like this
Squad seems to be very appealing for businesses that already have tractions
will double down on this https://t.co/pUEb1Nk8Vl
Simon Høiberg@SimonHoibergindie162.5K粉 · 8条讨论799美元软件包、一次性付费、自托管替代和高强度Agent订阅成本。
The awesome part of selling a $799 software bundle is that you don't need many sales before it starts getting fun.
I'm doing just a couple of sales/day. And that's $40K-$60K/month.
Then 2-3 times/year I do promotions which makes the revenue spike in those periods.
Convincing people to buy your stuff is hard. The value need to be crystal clear. But once you've convinced them, having them pay $799 instead of $120 is surprisingly little harder - at least for the right kind of buyer.
Generally, it's easier to convince 2 people to buy for $799 than 6 people to buy for $120.
引用@SimonHoiberg"How can you sustain a lifetime deal"
- It's a question I get over and over.
Here's how 👇
(FounderStack vs. AppSumo).
Our price is 15-20x higher than the average product on AppSumo - that's mostly it.
If you see software with ongoing costs offered at $39 … ↗
You don't really need DaVinci MCP.
Just let GPT 6 Astra run scripts and commands directly and combine it with computer use when DaVinci's API falls short.
The MCP actually gets annoying and a bit limiting/in the way.
Otherwise, I agree. This is how my videos are produced now too.
引用@gpumaxxerI’m spending my entire annual salary at Higgsfield on a billboard built like a Greek god.
Every ad slot on @marclou’s body is now Higgsfield. For HYROX.
Whoever wants to sell ad slots on their bodies, hit me up. @sydney_sweeney@KingJames 👀
fyi, my GPT-6 … ↗
Now showing in top-3 under "Omarchy" 😎
Been quite a while since I've appeared high in YouTube search for my videos. https://t.co/EADyX0ZfFP
引用@SimonHoibergOmarchy just completely changed how I'm building my AI-Native companies.
Let me show you the full breakdown here 👇
https://t.co/yr0XQDGQAS↗
I love this 👏
Here's a few I want to add:
- Buffer/Hootsuite -> FeedHive
- Intercom/Zendesk -> Aidbase
- Bitly/Shortly -> LinkDrip
- Sift/ClearSale -> SignupGate
Instead of vibe coding your own, get these replacements for a one-time purchase.
Save +$1000/m easily.
See here: https://t.co/sGIT7kWVMm
引用@levelsio✨ I replaced all these SaaS with my own vibe coded now, so about $25,000/mo savings:
- Weather API -> My own vibe coded weather service using Norway's meteorology API
- Image resizing API -> My own vibe coded image resizing API with NodeJS's Sharp + Redis … ↗
Every time I recommend self-hosted tools, people ask where they should rent the hardware.
For learning, testing, or running small, lightweight services - I recommend a small VPS.
The annoying part is that the price changes a lot by region. Storage, IPv4, bandwidth, taxes, and annual commitments can make the headline number pretty useless.
I went through the current starter plans and picked nine options with at least 2GB RAM, storage, and public IPv4.
Europe:
• Netcup - €5.72/month + setup
• OVHcloud - $5.35/month
• Hetzner - €11.99/month
US:
• RamNode - $5/month
• AtlanticNet - $8/month
• DigitalOcean - $12/month
Asia:
• BinaryLane Singapore - A$9.80/720 hours
• Vultr Tokyo - $10/month
• Linode Mumbai - $12/month
Before buying or renting hardware, I'd start here 👇
@thsottiaux - would love if you can make this happen. Happily paying $2000/m at this point to avoid rotating a bunch of subscriptions.
引用@SimonHoiberg@michael_kove Totally agreed! Or just a $2000/m plan with 200x (instead of 20x) so we don't have to manage 10 individual subscriptions. ↗
I'm spending 1x weekly limit per day with Astra. So I guess I'm gonna get myself 7x accounts (probably 8 to have a buffer).
In all fairness, $1400-$1600/m is still less than half of what I used to pay one full-time Upworker doing just one specific thing in my company.
My agentic OS is like a full team working 24/7.
Still worth it. Not even close.
引用@SimonHoibergAstra is too good.
So I'm adding a 5th (and probably soon a 6th) Codex subscription now.
But let's face it... We can't just keep stacking subs forever. It's not gonna get cheaper.
The rug pull is coming.
Start looking into local intelligence now. ↗
▸ 折叠2条(转推/噪音)
转推2026-09-10 RT @SimonHoiberg: I love this 👏
Here's a few I want to add:
- Buffer/Hootsuite -> FeedHive
- Intercom/Zendesk -> Aidbase
- Bitly/Shortly -…
转推2026-09-08 RT @SimonHoiberg: The single best use of AI right now is building your own internal tooling and custom harnesses.
What you need:
- Omarchy…
Jonathan Wilke@jonathan_wilkeindie28.4K粉 · 7条关注iPhone 18配色、Tailwind交易、CDN成本和语音驱动开发工具。
引用@rauchgYou can now go viral on 𝕏 without the stress of runaway CDN bills.
Crucially, without sacrificing our S-tier network performance.
AI is making devs and startups take off crazy fast. Flat Rate CDN brings you both scale and peace of mind. ↗
Finally got to work a bit on https://t.co/97kXWIBIpu today (I’m on vacation 🇦🇹) and I’m so excited about where this is heading 🔥
If this turns out well, I’m probably not going to use Cursor anymore. At least not directly.
引用@marclouI'm selling my body.
On September 19th, I'm competing in my first HYROX race in Turkey.
You can advertise your startup on 10 of my muscles:
- Your website is listed on https://t.co/49RnsFBVzP
- Your logo will be tattooed on my body on race day
- I'll share … ↗
have been trying to codify how i feel about technology, especially as i have more and more conversations with my kids about ai and the role of tech in their lives as teens/young adults.
they've just got entirely different perspectives/lived experiences and i've had trouble putting my finger on what to call those differences or the lens i view technology through.
initially my thought was "well, i'm a techno-optimist" but i don't know that i am. i don't really have an ideological view of technology in the "right and wrong" sense.
i'm just excited about technology. i see possibility. i see opportunity.
can technology be misused? obviously. just like all tools. but i don't believe they're inherently evil and i believe they are an incredible force for good when used appropriately.
so. what's that called?
in my head it's a lot like early radio. a real frontier. wide open spectrum, anybody with a rig could broadcast, nobody owned it. you could put a voice into someone's living room from a hundred miles away and nobody had to give you permission.
risk? sure. bad actors? plenty. unknowns? yup. but you could talk to strangers through the air. absolute magic.
and then it got licensed. networks bought the spectrum, and radio went from something you did to something you received. it's just what happens. every open frontier eventually gets fenced.
which is, i think, the actual gap between me and my kids. i got the internet while it was still spectrum. they got it after the networks moved in. i'm not more optimistic than they are. i just showed up earlier.
so what am i? i'm a frontier optimist. i thrive in the wide open spectrum.
about to simplify pricing on https://t.co/rHttZyJqV0
running a store is a constant whack-a-mole battle against bad actors stealing your brand, your photos, your name and everything in between.
knockoff fights that battle for you so you can keep growing your empire. the end. https://t.co/kftu1lafIi
not to go grandpa mode here, but something i notice w/ a LOT of younger generations of entrepreneurs is total comfort in saying literally anything in the name of marketing.
no qualms about making baseless, impossible claims and guarantees.
this WILL come back to bite you.
I'll say this again: Anthropic is the worst company in all of history at PR.
引用@EvanHubJacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not … ↗
If you've been hesitant to sign up for an Initial Commit Club membership, you can now get fully refunded so that you're account is free!
For each person you refer, you get 1/3 of the cost of your account refunded to you. 🎉
https://t.co/retNqhtJ21https://t.co/Aupds9CQ04
babe wake up
it's 2036
you need to use grok mega pickle 12.1 for 90 minutes to pay for your chipotle dinner last installment
the future is bright https://t.co/488nMwseXu
引用@Mutchtaba2STOP PAYING FOR TOKENS !
START GETTING PAID INSTEAD
Introducing Straitly, the first AI service that PAYS YOU for using tokens
Unlike
> Openrouter charges 5.5% service fee
> Vercel charges 2.9% payment fee
> Anthropic gives you low rate limits
> OpenAI only … ↗
new in activationpal
> last 14 days / custom range
> trial churn in key metrics
> faster loading lol
> trial insights per app
> expand charts
> click on any data, to see users journey (forgot to record)
next big feature — retention insights https://t.co/V8mAg1G7d8
Robots getting tested inside a browser tab still sounds made up
→ no rig booked
→ no lab time
→ no shipping a robot across a country
Antioch from @antiochrobotics runs the whole physical AI loop from a tab, then hands back hours instead of entire weeks
引用@crestaWe’re welcoming Super Bowl champion, Bay Area legend, and media icon @MoneyLynch to Team Cresta.
He earned the name Beast Mode by being unstoppable. Now your customer experience can be too.
Call the right play. Solve the problem. Never fumble.
Put your … ↗
转推2026-09-09 RT @DanKulkov: new in activationpal
> last 14 days / custom range
> trial churn in key metrics
> faster loading lol
> trial insights per a…
转推2026-09-09 RT @DanKulkov: https://t.co/zPHYwtMA85 is growing
product & revenue analytics for ios apps 📱
free plan available
using it myself to grow…
引用@marckohlbruggeI feel like the “Apple Reference Image” feature is a sleeper
Imagine an Instagram clone where only verified unedited images can be posted. Same for dating apps, etc
This will bring back trust to social media in a time where generative AI has eroded it to a … ↗
I feel like the “Apple Reference Image” feature is a sleeper
Imagine an Instagram clone where only verified unedited images can be posted. Same for dating apps, etc
This will bring back trust to social media in a time where generative AI has eroded it to a large extent
Also opening up a market for completely new ideas we haven’t even conceived yet
引用@marckohlbruggeApple announced this just now 🔥
It's called "Apple Reference Image" and comes with iOS 28 ↗
iphone duo app idea:
game where you need to fold and unfold it as often as possible in 10 seconds
compete with friends
who can fold their phone the most?
Apple announced this just now 🔥
It's called "Apple Reference Image" and comes with iOS 28
引用@marckohlbruggeIs there camera technology that can somehow include an authenticity token in photos/videos to prove photos are genuine and untampered? ↗
引用@GergelyOroszSince SpaceX acquisition of Cursor:
Cursor Community --> renamed to SpaceX AI Community
Cursor Ambassadors --> renamed to Grok Bot Ambassadors
Café Cursor --> Cancelled
Very clear that SpaceX is retiring the Cursor brand, pushing everything to be … ↗
引用@OpenAIWe’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 … ↗
转推2026-09-09 RT @levelsio: I really liked @marckohlbrugge's e-ink screen by @itopinion
So I asked Claude Code to make a CSS UI Kit I can use for my si…
转推2026-09-09 RT @buildinpublic: today we're launching DEALS 🎁✨
630 deals on tools you will actually use worth over $3.3 million combined
→ Price: FREE…
引用@dannypostmagenuinely triggered to buy a foldable phone just for the larger screen real estate + multi-tasking to make this much better experience on the go ↗
引用@dzhngIntroducing Fly Escape Room - the world's first game with NPCs powered by a real fruit fly brain, all 166k neurons.
Leave food and other objects for the flies, run the simulation, watch what happens. Goal is to get all flies to escape.
Built w/ GPT-6 Astra … ↗
added dungeons and parties 🔥
dungeons can now be found hidden around the world, making exploring more rewarding if you find a hidden lair
queue up with your friends or village and slay the monster to get high-tier loot 🕷 https://t.co/ttwPyo0iTo
Jon Yongfook@yongfookindie171.5K粉 · 4条围绕iPhone Duo快速做个人站和折叠交互点子，也加入X新创作者计划。
I watched the Tetris movie last night. Brought back some childhood memories... so this morning I vibed a new personal site using Fable.
https://t.co/qbq3AgYHhJ
引用@hboonApple likes apps that show off new hardware. How about an iPhone Duo game where the primary mechanic is opening and closing it, say to squash a fly? ↗
Reverse trials are having a moment. Unlock everything then drop people to free and let loss aversion do the work.
A founder I worked with tried it. Conversion barely moved.
The reason was timing. New users didn’t need the advanced stuff yet. Those features solved problems that show up months in. So when the downgrade came nobody felt anything missing.
You can’t take away something people never started using.
反向试用失败的关键是高级功能价值出现得太晚。
Peter Askew@searchboundindie26.2K粉 · 3条记录Vidalia一个产季结束后的现金结算现实与求职者素材。
so I essentially work without pay on Vidalia for the whole season. once we wrap up, I pay all our remaining bills and see what's left. if there's nothing, then homey gets zilch.
X ads used to be terrible, but they seem very compelling now. You can target the followers of accounts you handpick, and you often get a cost per click in the 6c-10c range. Also, it seems like engagements from ads boost organic reach. For example, the ad may bring 1M impressions, and the same post gets another 2M impressions organically, presumably boosted by the engagements from the ad. I've seen this happen many times in the past 4 months.
Some people try to reduce the ads they see on the platform by blocking every account they see running an ad, so there is still a downside that you get mass blocked by thousands of people. But it might still be worth it if you want to get a spike of attention to something important you want to promote.
引用@elonmuskIf you want to reach the smartest people in the world, advertise on 𝕏 ↗
"Consulting is bad, only sell products."
Do both.
I wanted only to sell products, but then something strange happened.
Some customers wanted additional value and had $20k, $30k even $50k budgets.
It would have been insane to turn those customers away.
"Sorry mate, I only sell products"
If you have experience and knowledge that is highly valuable, you're doing yourself a disservice by not packaging it.
Only indie hackers have this narrow-minded mindset.
Every major SaaS company offers professional services to enterprise customers.
The clue is in the second S of SaaS.
If you want a successful SaaS you have to do the BORING work.
You can’t just post screenshots on X all day.
Do SEO, cold email, LinkedIn.
If you don’t want to do those things, don’t start a SaaS.
AI coding tools will confidently suggest changes that will piss off your entire customer base.
Today I was re-assessing how Alertly allocates social monitoring budgets across the month.
Astra's solution: remove the budget allocation entirely and force everyone who hits limits early to upgrade.
Sounds great - revenue expansion right?
Except customer conversations scattered across DMs, calls, emails and support tickets tell a different story.
When customers hit limits too early, it causes confusion, frustration and leads to churn.
AI doesn't read between the lines of old customer conversations.
It doesn't ask "how does this impact the existing customer base?"
It just sees a technical problem and outputs a technical solution.
This is where experienced operator judgment matters.
You need to know:
> How many customers this impacts
> What their usage patterns look like
> Whether to grandfather or migrate
> What the support burden will be
> If the change is worth the friction
If you don't ask the right questions, you're putting your customers and revenue at risk.
The solution I went with was to maintain the existing budget allocation but improve the feedback on usage and introduce a softer upgrade path.
A business decision, not a technical one.
I'm writing the first lessons for https://t.co/cAOrzKSqxk today.
People balk at selling courses, but I love making products that help others succeed.
If even one person applies this stuff and gets paying customers that's a huge win for me.
Applied for the new x monetisation programme, got approved in < 1hr
Interested to see what the payouts will be compared to the old one. https://t.co/mRzr5vZAiu
Chasing status is a hamster wheel where you live your life based on other people's expectations at the expense of your own happiness.
If your self worth is tied to MRR or leaderboards or what strangers think about you on X then you are cooked.
There will always be a new milestone, someone to compare yourself to.
You will never arrive.
The irony is no one cares how successful you are or how much money you have.
You can spend your whole life trying to win a game that makes you unhappy.
Or you can just opt out.