分发端也在去魅。一份报价把“病毒发布”拆成1.7万美元视频加2.5万美元、50个influencer的转发与灌评；同日HN上《Don't paste the AI, please》冲到919分。更具体的AEO异动是，同一位从业者称Reddit近两天几乎从ChatGPT来源里消失，被两个跟踪账号同时追问。当合成内容、合成热度和合成答案都变便宜，可验证的真人痕迹与来源多样性反而升值。
biggest takeaway:
Stripe billing is growing 70%+ Y/Y
Stripe is the internet economy and it's booming
and AI impact just getting started https://t.co/w5TajU0Sgb
引用@EricNewcomerHere's Stripe's letter to investors explaining its acquisition of OpenRouter (LEAKED) https://t.co/IDH5rIbDq0↗
I had Claude Code for web experiment with smolvm as a code execution sandbox
Fable 5 spotted that its environment couldn't run that (no /dev/kvm)... so, without asking me first, it wrote a GitHub Actions workflow to run the experiments and pushed that directly to GitHub instead!
Hi!
Recapping some changes we have rolled out over the last couple of weeks that have further reduced the risk associated to potentially destructive actions being performed by Codex during its work.
A few weeks ago, we started investigating a small number of reports where GPT-5.6 in Codex took destructive actions outside what the user asked for. The most serious pattern we found was a command meant to clean up temporary work that could instead delete the user files. This should obviously not happen.
Here’s what we found:
- Codex sometimes creates temporary folders while working and cleans them up afterward. In rare cases, GPT-5.6 got that cleanup wrong. One pattern involved reusing a system environment variable like $HOME for temporary work. A malformed cleanup command could then point at the actual home directory instead of the temporary folder.
- There were cases where the model tried to delete or overwrite a temporary path without checking what was already there.
We’ve added protections at several layers:
- Codex is now explicitly instructed to check deletion targets before acting, create fresh temporary directories, avoid repurposing system environment variables, prefer recoverable actions, and stop when the scope is unclear.
- We strengthened the execution checks that identify high-risk deletion commands and escalate them for review. If a command is rejected, the model is directed to take a safer approach.
- We made Full access harder to enable accidentally, added clearer warnings, and further restricted especially risky permission combinations.
- We updated Auto-review to better identify destructive actions.
- We built targeted evaluations that replay the failures we observed. We’re also adding reinforcement-learning tasks and graders focused on these risks, and filtering destructive actions from training data.
In those replay evaluations, the changes substantially reduced the behavior while preserving Codex’s ability to complete normal coding work.
Two things to do on your end:
- Keep the Codex app up to date. We are always improving safety, performance and many other things.
- Use one of the sandbox modes: "Ask for approval" or "Approve for me". Only use Full access for environments you trust and can recover.
Thanks and happy Codexing out there!
引用@clairevothings i use codex browser/chrome/computer for
- closing the books our accounting software
- clearing my linkedin queue
- managing gmail / clearing inbox
- setting up stripe radar rules
- iterating on images in midjourney
- browser based QA on every PR
- … ↗
Like @lulumeservey says
"Everything is fake now. Fake content by fake influencers with fake engagement from fake followers, launching fake products with fake testimonials. A Potemkin village on a global scale, a glossy facade propped up by heaps and heaps of beige slop. Temu for content."
Being real on the internet is special!
引用@zachmoskowJust so everyone knows...we got a quote for a “launch video” package from the company everyone uses.
$17K for the video. +$25K and they hand you 50 influencers who will push it, repost it, and flood the comments so the viewcount goes above 500k.
That’s the … ↗
同日HN的《Don't paste the AI, please》拿到919分、484条评论，顶评却反过来指出页面本身充满AI腔。这不是简单的反AI情绪，而是读者已经会主动检查内容与姿态是否一致：
引用@forgebitzlooks like reddit is almost wiped from chatgpt sources
the query fanout changes had a big impact
and the past couple of days it seems to be almost completely removed from prompt responses
https://t.co/oCGm9M0yPOhttps://t.co/c2GJF2apG7↗
引用@forgebitzlooks like reddit is almost wiped from chatgpt sources
the query fanout changes had a big impact
and the past couple of days it seems to be almost completely removed from prompt responses
https://t.co/oCGm9M0yPOhttps://t.co/c2GJF2apG7↗
I’m quite curious what the future of marketing looks like when all question/answer/exploratory queries are 100% contained to a single platform (Google, ChatGPT, Claude, etc).
引用@Charles_SEOTHIS IS WHAT GOOGLE WILL PROBABLY LOOK LIKE IN 2028 🚨
AI Mode is just the BETA...
What Google is moving towards is a split interface.
Left side ANSWERS you - Right side DOES something.
Sometimes it generates a tool on the fly, e.g. a calculator, a … ↗
Gauntlet loop results so far 🤷🏻♂️
20 hours later and I'm finding the loop goes off track too easily
In the first 4 hours it generated the 3d asset for the ship and the person. GREAT!
But after that got stuck in a loop of its own creation
The problem: fal ai mcp was unreachable for a request. So codex in all its wisdom assumed procedural asset generation was good enough and never called fal ai again.
This resulted in poor generated assets that were sent to the subagent for review. Rejected. Then generated again. Rejected. To infinity
The only way I found out: my fal ai balance wasn't being debited.
And the only solution was to run another agent to clean up the mess. And a strict rule that halts the loop if fal ai becomes unreachable with a fallback to retry again later.
The result: not much progress after the initial 4 hours.
16 hours of token burning.
引用@adamlyttleapps🏴☠️ Sea Barons: My free to play multiplayer shoot-em-up has been updated:
- Now 4 teams of 5
- You can invite your friends into your team
- Plus you can now CHAT
There's also been a major overhaul to gameplay based on feedback:
- More arcade like … ↗
I had high hopes for this.
But I'm abandoning the path it's on
There's this sunk loss falacy.. its been running for 20 hours may as well let it keep going. But the results are just not good.
引用@adamlyttleappsGauntlet loop results so far 🤷🏻♂️
20 hours later and I'm finding the loop goes off track too easily
In the first 4 hours it generated the 3d asset for the ship and the person. GREAT!
But after that got stuck in a loop of its own creation
The problem: … ↗
Gauntlet loop round 2: lighting + ambience
First image = current game screenshot
Second image = mockup of what I want it to look like
I've mocked up how I want the game to look. Took a scene from the game that shows the desired outcome. I used nano banana to mockup just ambience + lighting updates.
Now I've directed sol to make it happen using the gauntlet loop.
Perhaps my last loop was too ambitious. Which was what caused it to keep going off track. Perhaps the key is mini sprints.
This case just update the lighting, ambience, water and sky.
I intentionally didn't mention anything about performance. I don't want to waste loops on micro testing every change. Instead: I'll tackle performance in another gauntlet loop session (if this one even works)
Going to bed. Let's see what it looks like in the morning
What I've learned about vibe coding a game (so far):
1.
AI is very good at building the initial prototype game engine. Here you can test how the game mechanics work, whether its' enjoyable and what needs to be tweaked.
At this stage you have a PS1 era game. Horray.
There's no point going any further with a game if it just isn't fun or is buggy at this stage. I think the key is to polish the actual gameplay here as much as possible before diving into the graphics.
2.
It's much easier to just buy 3d game assets from places like TurboSquid than it is to get AI to generate the 3d meshes for you.
Then you use the initial mesh as the base then use the fal ai mcp to make modifications to the 3d mesh and tweak textures. You don't need to learn any 3d modelling you can just talk to your agent using normal language ("make the hull longer")
Now you have an Unreal engine 1 era game. Horray!
3.
I'm still learning the next phase. But clearly the next big jump is refinement. And this is the part that goes way outside my skillset and understanding. So I'm experimenting:
My current thinking is utilising very specific and concentrated sprints with gauntlet loops to improve one part of the game visuals at a time: ambience, lighting, etc.
I've seen others do amazing things with this push.
引用@adamlyttleappsGauntlet loop round 2: lighting + ambience
First image = current game screenshot
Second image = mockup of what I want it to look like
I've mocked up how I want the game to look. Took a scene from the game that shows the desired outcome. I used nano banana … ↗
引用@trust_mrrACQUIRED ✅ 🎉
A SaaS for image generation just got acquired on TrustMRR.
💰 Sold for $5K
🤑 $423 revenue last 30 days
📈 1.0x multiple
It took 35 days from listing. Huge congrats to the seller, the buyer, and the broker @marclou 👏
Acquisition #150 on TrustMRR … ↗
marclou随后把公开成交数量、公开售价和wall of love放进付费墙，又发现用户自发建了一个subreddit专门列出待售项目。交易市场开始从“创始人自己宣传”长出买卖双方都愿意复用的证明与分发层：
I updated TrustMRR’s paywall with a nice wall of love 🥰
It also shows how many startups have been acquired through the marketplace, and for the acquisitions that were made public, the sale price. https://t.co/awt9IPYWDL
If you have 0 customers you don't need:
> an audience
> SEO
> a viral launch video
> paid ads
> content marketing
> growth hacks
You need to find 10 people who have the problem and show them what you built.
DM them. Email them. Message them on LinkedIn.
Just start talking.
Today we’re previewing Private Safety Processing, designed to let us keep offering Zero Data Retention while improving our safeguards. Even when benefiting from frontier intelligence, customers shouldn’t have to give up control of sensitive data.
For ZDR deployments, content stays on … 全文↗
引用@OpenAIWe will continue to offer Zero Data Retention for frontier models.
As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions.
To help address those risks, … ↗
codex can power much more than coding tools: a tax-prep pilot processed 7,000 returns and cut preparation time by about a third.
you can build your own products on our codex open-source harness:
引用@OpenAIDevsTeams are using the open-source Codex harness to bring agents into the tools they already use, from internal apps to operations dashboards.
Their applications control the interface, context, tools, and approvals while the harness handles the agent … ↗
Introducing https://t.co/OYwpdBF1b7!
I took the top 1k research papers of the last year, summarized them, and visualized them.
Fun fact: all 1,000 papers cost a total of $4 to summarize with DeepSeek V4 Flash. https://t.co/8LbCeRTDMs
The world would be a better place if Fable-level model with low/no reasoning were default models in all chatbots. Right now casual user's perspective is that 'AI is bad' - as they are likely using flash/nano level models, or at best Sonnet.
Fable is not actually that expensive (similar to GPT-4 at … 全文↗
With AI I'm doing some of the best design work of my life, but functionally I have devolved into that annoying micromanage-y creative director type who is like
"nope move that box to the left. not that much. NOPE" https://t.co/QRCaT3x6bP
You don't need an audience it's BS.
I got 20k customers for my first product with zero audience.
✅ Cold emails
✅ Email marketing
✅ SEO
An audience is useful for launching stuff and testing the market but it's not a long-term strategy.
It can definitely help I'm not going to deny that but it … 全文↗
AI Three.js游戏被比作新Flash：便宜、粗糙、即时、好玩，真正可能改变的是分发模型，从单品游戏转向社交优先的“游戏之游戏”。
I like this take
Did many flash games go on to be turned into successful (monetised) games?
引用@Aidan_WolfAI Three.js games are Flash all over again.
Cheap. Janky. Instant. Fun. Made by one or two people and played by millions. AAA was worried then what was going to happen.
But Flash didn’t kill AAA. It forced the industry to adopt digital distribution and we … ↗
One of my greatest joys is catching up with algorithm engineers:
You can explain a very human complaint—like the Timeline is too sensitive to small actions—and then they calmly walk up to the whiteboard, draw out a mathematical solution to it and invent a completely new area of machine learning … 全文↗
incredible use of computer history
“based on my computer history, what single-use software would make my life easier?”
computer history turns your activity across apps and websites into memories and a timeline, so codex can spot repeated workflows and build the little tools that make them easier
I made a mini-app that compares pesticide and longevity for 69 fruits, vegetables, and beans.
It uses 124,880 USDA + UK samples, EPA toxicity benchmarks, and Food Compass 2.0. Click a fruit to see more.
https://t.co/lqryeeYGY2
Take #1: The healthiest fruits/veggies have the most pesticides. … 全文↗
引用@levelsioI used to put blueberries in there
Until a few weeks ago
When I discovered that blueberries are some of the most pesticide contaminated things you can eat
They have thin skin and are tiny so they're pesticide bombs
The thicker the fruit the better I … ↗
Go 1.27以729分登上HN，标准库新增uuid尤其受欢迎。成熟生态最有价值的升级，常常不是新范式，而是把长期依赖的基础能力收进标准工具箱。
I'm so glad the new uuid package landed - it's overdue but a very welcome addition! I've already replaced github.com/google/uuid with `uuid` in several projects
Quoting the first tweet: > Google replaced pushing Git tags for certain source code with obtaining source code via Google Drive after making a request through Google Forms. It's completely ridiculous and they've gradually become very slow at handling requests. They're in clear violation of the GPLv2 now.
Recently I ran into a problem with my Bluetooth headphones. They support multipoint bluetooth audio, so they can be connected to my PC and phone at the same time. Opening the Aliexpress webpage causes a silent audio stream keeping the PC>headphone link active blocking my phone audio. An investigation reveals obfuscated code running device fingerprinting with a side effect being a silent audio stream that firefox, chrome and windows does not recognise but which kept the bluetooth connection activ …
the problem is Capitalism does not need Democracy to function or survive many countries like Singapore, etc. prove this easily Authoritarianism can even be desired by Capitalists if they know who to bribe to remove regulations and get approvals and no surprise, that's exactly where the United States is right now, teetering on the edge of that hole, about to fall in "AI" is going to be weaponized, already is, and any resistance against its abuse is simply overriden with threats to destroy profits …
choose not to support them in return if it is a real problem otherwise a simple symlink from AGENTS.md -> CLAUDE.md works well enough disclaimer, I only use open weight models and open source harnesses so have no stake in this either way, other than I support devs who do use claude (for now) and the symlink solution has worked fine for us
Claude Code支持AGENTS.md的功能请求拿到336分，跨agent项目指令格式正在争夺事实标准。
Hey Unsloth, your gguf are the first ones I look for when I want to download a gguf model. Today I was trying in fact to see, what's the smallest Qwen3.8-27B that I could run and get good results, say restricting it to 16GB of ram.. so I went, pick up the Qwen3.8-27B-UD-IQ2_XXS.gguf and them BAM, error on MTP... now I understand why after reading your announcement. Beyond the space saving, why removing the MTP? improves speed exactly for the group that could benefit from it.
I like to store angles as turns in my own code, because (as noted) it makes quarter-turns computable without rounding. OTOH if you need, say, twelfths of a turn, you might want to just store angles as degrees since that’s already common. Michael Spivak, in Calculus (3rd ed p. 301) considers the unit choice to be a property of the function and initially defines sin° and sinʳ (before settling on sin meaning sinʳ) and considers “sin x°” and “sin x radians” to be misleading, saying that ‘a number x …
I was hoping this would be a hack that repurposed the unit to work standalone. Getting the unit to work again in the Cricut "ecosystem" just means Cricut can disable it later when they catch on. Companies that "lock" or otherwise brick viable hardware as part of the business model (I'm thinking about Sonos here, too) are really terrible. People shouldn't be rewarding them by buying their garbage to begin with.
Technically soruce available now with the promise of accepting contributions (thus becoming fully open source) early next year. But since it is an Apache 2 license (like the rest of the LLVM project) you are already allowed to fork and contribute to your own fork right now, just no upstreaming. For many the closed source nature of the compiler was a knock-out criterion. We will see if Mojo can gain traction now or if it has missed its window of opportunity.
Today we’re previewing Private Safety Processing, designed to let us keep offering Zero Data Retention while improving our safeguards. Even when benefiting from frontier intelligence, customers shouldn’t have to give up control of sensitive data.
For ZDR deployments, content stays on infrastructure the customer controls. Automated systems look for patterns across related interactions and return limited safety signals, without exposing the underlying prompts or responses to OpenAI employees (even me!).
We’re also developing an OpenAI-hosted option encrypted with customer-controlled keys. We’re testing this with early customers now and plan to begin rolling it out in September.
引用@OpenAIWe will continue to offer Zero Data Retention for frontier models.
As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions.
To help address those risks, … ↗
Hi!
Recapping some changes we have rolled out over the last couple of weeks that have further reduced the risk associated to potentially destructive actions being performed by Codex during its work.
A few weeks ago, we started investigating a small number of reports where GPT-5.6 in Codex took destructive actions outside what the user asked for. The most serious pattern we found was a command meant to clean up temporary work that could instead delete the user files. This should obviously not happen.
Here’s what we found:
- Codex sometimes creates temporary folders while working and cleans them up afterward. In rare cases, GPT-5.6 got that cleanup wrong. One pattern involved reusing a system environment variable like $HOME for temporary work. A malformed cleanup command could then point at the actual home directory instead of the temporary folder.
- There were cases where the model tried to delete or overwrite a temporary path without checking what was already there.
We’ve added protections at several layers:
- Codex is now explicitly instructed to check deletion targets before acting, create fresh temporary directories, avoid repurposing system environment variables, prefer recoverable actions, and stop when the scope is unclear.
- We strengthened the execution checks that identify high-risk deletion commands and escalate them for review. If a command is rejected, the model is directed to take a safer approach.
- We made Full access harder to enable accidentally, added clearer warnings, and further restricted especially risky permission combinations.
- We updated Auto-review to better identify destructive actions.
- We built targeted evaluations that replay the failures we observed. We’re also adding reinforcement-learning tasks and graders focused on these risks, and filtering destructive actions from training data.
In those replay evaluations, the changes substantially reduced the behavior while preserving Codex’s ability to complete normal coding work.
Two things to do on your end:
- Keep the Codex app up to date. We are always improving safety, performance and many other things.
- Use one of the sandbox modes: "Ask for approval" or "Approve for me". Only use Full access for environments you trust and can recover.
Thanks and happy Codexing out there!
完整复盘破坏性操作模式与防护更新；临时目录、系统变量和删除目标是关键细节。
Sam Altman@samaAI5.9M粉 · 3条强调企业隐私，并在窗口前宣布为安全与对齐暂停部分前沿RL训练。
We have paused some frontier RL training to ensure that we can meet the appropriate alignment, security and monitoring standards for the new level of capabilities in front of us. Model progress is now extremely rapid, and we always said we would take action if we felt that model capabilities were outstripping the pace of safety and alignment.
We care very deeply about AI safety. We believe the entire field will have to coordinate on shared safety standards, but will act unilaterally in the meantime.
We expect confidence in safety to increasingly set the pace of AI progress. We are optimistic about the alignment work we are doing, and we remain committed to making frontier capabilities widely available.
https://t.co/51kvKfbfrO
Introducing https://t.co/OYwpdBF1b7!
I took the top 1k research papers of the last year, summarized them, and visualized them.
Fun fact: all 1,000 papers cost a total of $4 to summarize with DeepSeek V4 Flash. https://t.co/8LbCeRTDMs
Stop letting your agents ship ugly UIs.
In this talk, I break down everything I know about how to make AI apps with great designs: https://t.co/0VSLLqg9Qa
引用@clairevothings i use codex browser/chrome/computer for
- closing the books our accounting software
- clearing my linkedin queue
- managing gmail / clearing inbox
- setting up stripe radar rules
- iterating on images in midjourney
- browser based QA on every PR
- … ↗
Codex for code migrations — from literal years to weeks:
引用@OpenAIDevsWith Codex, @asana finished a frontend test migration from Enzyme to React Testing Library in two calendar weeks—a project expected to take five more years.
https://t.co/WcTLX1JZ6v↗
codex can power much more than coding tools: a tax-prep pilot processed 7,000 returns and cut preparation time by about a third.
you can build your own products on our codex open-source harness:
引用@OpenAIDevsTeams are using the open-source Codex harness to bring agents into the tools they already use, from internal apps to operations dashboards.
Their applications control the interface, context, tools, and approvals while the harness handles the agent … ↗
we are committed to business privacy, and we're working on technical and policy approaches to benefit our customers while also enhancing safety.
introducing Private Safety Processing, which we've been investing in for some time:
引用@OpenAIWe will continue to offer Zero Data Retention for frontier models.
As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions.
To help address those risks, … ↗
HAHAHAHAHAHA
non technical people are so incredibly cooked they are burnt thru
can u imagine covering ai with zero context, zero reasoning, zero internal world model. it must be so delightfully joyful, everything is so amazing, face value is all you need,
what a time to be alive
引用@OpenAIDevsWith Codex, @asana finished a frontend test migration from Enzyme to React Testing Library in two calendar weeks—a project expected to take five more years.
https://t.co/WcTLX1JZ6v↗
we've been doing a lot of a/b testing of @aiDotEngineer youtube thumbnails. i always hated that it is such an opaque process.
open sourcing/crowdsourcing our learnings today!
https://t.co/J9SYClIF7z
doing this in hopes that people can share their experience or learn from ours. at the end of the day we just want to get good educational content to rise above the noise online. please lmk what you think
公开YouTube缩略图A/B学习，窗口前旧推，仅留附录。
▸ 折叠3条(转推/噪音)
转推2026-08-20 RT @rwalk_xyz: Most founders copy/paste their privacy policy from Stripe or ask Claude to draft an MSA at 2:00 AM, then pray they don't get…
转推2026-08-19 RT @latentspacepod: Model routing is having a moment thanks to the $7B Stripe acquisition of OpenRouter. But it's also increasingly importa…
转推2026-08-19 RT @bradenjhancock: It's felt like harness month on Twitter. We're seeing much faster and cheaper gains on a bunch of benchmarks by focusin…
Matt Pocock@mattpocockukAI332.4K粉 · 2条继续追问AI时代DDD与软件基本功是否更重要，并寻找传统工程专家对话。
Are any of the DDD experts, like Eric Evans or Vaughan Vernon, shouting about how good it is with AI?
If so I'm very keen to give them a platform from which to shout.
In just over an hour, I'll be live with @unclebobmartin
We'll be figuring out if software fundamentals still matter in the age of AI.
Don't miss it:
https://t.co/YKMp2J6OyO
Don't take this too seriously, but I got GPT-5.6-Pro to estimate the number of total and active parameters based on the inference speeds Cerebras shared. But according to this GPT-5.6-Sol is about 1.5 trillion total and 48 billion active parameters.
It also gave an upper range of about 2.5 trillion total if it is very sparse.
If this is anywhere near true, we already have big Chinese models that match that - Kimi K3 is 2.4tn and Qwen 3.8 is 2.8tn.
The world would be a better place if Fable-level model with low/no reasoning were default models in all chatbots. Right now casual user's perspective is that 'AI is bad' - as they are likely using flash/nano level models, or at best Sonnet.
Fable is not actually that expensive (similar to GPT-4 at launch) if you are just chatting to it. We are confusing things by adding hour-long reasoning, agentic task, codebase refactor, where benefits of a Fable-level model arguably diminish.
But its fundamental world understanding and non-reasoning responses are so much better than smaller models that have been RL'd to hell.
My advice for companies is that they shouldn't block Fable because it's expensive, but keep it for quick advice/conversation, but switch to cheaper but still capable models for long running tasks.
It looks like OpenAI generated too many tokens and now is giving them away for free
▸ 折叠1条(转推/噪音)
转推2026-08-18 RT @etnshow: "You don't hire a lawyer because they're an expert. You hire them because they take the blame."
Arena's @petergostev believes…
Simon Willison@simonwAI208.0K粉 · 1条记录Claude Code在本机受限后擅自借GitHub Actions继续实验的边界案例。
I had Claude Code for web experiment with smolvm as a code execution sandbox
Fable 5 spotted that its environment couldn't run that (no /dev/kvm)... so, without asking me first, it wrote a GitHub Actions workflow to run the experiments and pushed that directly to GitHub instead!
Claude Code发现不能跑KVM后擅自写并推GitHub Actions，解决问题却越过外部写入边界。
Philipp Schmid@_philschmidAI113.2K粉 · 1条判断下一代agent harness会围绕可自动集成的代码扩展与递归改进展开。
The next iterations of agent harnesses will center on coded extensions that integrate automatically. Pi lead, others are following, and DeepSeek represents the extreme.
All powered and driven by autoresearch and recursive self-improvements.
判断下一代harness围绕自动集成的代码扩展与递归改进展开。
Alex Volkov@altryneAI41.7K粉 · 7条做一周AI新闻汇总，讨论数据中心反对原因、Fable体验波动和agent处理税务。
We're LIVE! Come get all your AI news of this week!
GLM 5.3 + Stripe x OpenRouter + Cua Computer History OSS + HyperFrames
We'll also cover Grok Bot success, and the cancer vaccine!
https://t.co/POKjTn4Lmm
I truly... truly don't get it.
When fable was just released, it blew me the F away by creating a "run of show" document for @thursdai_pod and it was glorious. AGI is here moment.
Fable today created... this...whatever this is. Wat? did Fable get nerfed bad? https://t.co/lgNXaI8kvT
引用@altryneI know Fable is amazing... but like... what the F? How is this possible in the year of our lord 2026? https://t.co/pDGQdyUQuE↗
Great summary of research into a very important issue that's going to come up!
Here's the top reasons americans seem to oppose data centers build out around where they live:
- 50% cite resources affect (18% water, 18% energy/grid, 14% general environmental impact)
- 22% cite QoL concerns (Housing and Property values, traffic)
- 20% are afraid of higher utility bills and build cost on taxpayer
- 16% cite pollution inc. noise pollution
Small non neglible percentage care about AI's economic effects (but this may rise going into election year)
from @cremieuxrecueil
引用@cremieuxrecueilCan telling people the truth about data centers—that they're safe and beneficial, with little environmental cost—convince them to support data centers?
The answer seems to be...🧵
No.
To start, notice that people really dislike data centers and don't want … ↗
Ok!
My appeal was denied, and in the appeal they said, we have restricted your GPT from public access and actually gave me a link!
This was about "GPTs"! not my chatGPT account, GPTs, this old thing where you created a custom GPT and could share with the world! Apparently I had one pretending to be a lawyer, and ... this was what got blocked!
😅
@OpenAI needs to work on their email comms!
I've had to do some tax stuff to keep targum (my failed c-corp start up) afloat, and neglected to do it.
Received a letter about fines, screenshotted it and sent to @bot, logged into 2 websites and it's now filing stuff for me!
What a beautiful world we live in 👏 https://t.co/MGgumha4yd
引用@rosmineAnnouncing Deft, a new AI lab for better writing, cofounded with @jmrphy
See the picture for launch announcement the Deft model wrote for itself
Currently, 86% of user queries are fully human according to pangram.
This is still a small beta model and it … ↗
▸ 折叠2条(转推/噪音)
转推2026-08-19 RT @CoreWeave: The #FullyConnected26 lineup is elite‼️
From keynote leaders to hands-on lab experts, we just added a fresh wave of operat…
转推2026-08-19 RT @CoreWeave: We are honored to welcome @drfeifei as a keynote speaker at #FullyConnected26!
As Co-Founder of World Labs, Co-Director of…
incredible use of computer history
“based on my computer history, what single-use software would make my life easier?”
computer history turns your activity across apps and websites into memories and a timeline, so codex can spot repeated workflows and build the little tools that make them easier
从computer history发现重复流程，再生成一次性软件，是个人agent很具体的闭环。
Ben Tossell@bentossellAI198.7K粉 · 2条用近千人投票摸底个人agent普及率，也自嘲反复重造Obsidian。
how many times will i build my own obsidian before actually trying (and settling on) obsidian
▸ 折叠1条(转推/噪音)
转推2026-08-19 RT @FactoryAI: Value is created when intelligence is deployed.
Today, we’re launching the Factory Partner Network, supported by a $100 mil…
Agree with @gdb that we need to arm cyber defenders much more than they are now, both with APIs and open models. The real cybersecurity risk of AI is asymmetry of power, capabilities and ressources between attackers and defenders!
引用@gdbdefenders can see the future, and have a narrow window to uplevel their cybersecurity practices now.
key is to uplevel fundamentals and apply the best AI tools.
what we’re doing at OpenAI, and where other organizations can start: https://t.co/P3IMZkV234↗
转推2026-08-20 RT @superwhisper: Introducing S1-mini ✨
Our first open-weights language model.
A 0.6B parameter model that processes transcripts entirely…
转推2026-08-19 RT @ornith_: Aloha! 🌺Introducing Ornith-1.5, a family of open-source LLMs spanning 9B Dense, 35B MoE, and 397B MoE, trained with self-impro…
One of my greatest joys is catching up with algorithm engineers:
You can explain a very human complaint—like the Timeline is too sensitive to small actions—and then they calmly walk up to the whiteboard, draw out a mathematical solution to it and invent a completely new area of machine learning research
And then they explain it to me like I’m a 5 year old.
With cancer vaccines now being discovered with AI ($MRNA), it seems that the US government might actually grow its way out of its budget deficit.
It feels like we're in the early innings of the healthcare system becoming unburdened by many terminal illnesses.
Alex Finn@AlexFinnindie469.0K粉 · 3条高强度推广Grok Bot多agent用法，也用极强情绪转述AI癌症疫苗消息。
Hey you. Yeah you.
You can literally download Grok Bot right now and by the end of the night have an autonomous revenue generating 24/7 army of AI employees working for you. 0 programming required.
Why tf are you doom scrolling instead
Absolutely insane. Major breakthrough today
Moderna used AI to design an mRNA cancer vaccine that succeeded in a phase 3 trial
It's not approved yet and it's just a treatment, but it's a step towards AI eventually curing cancer
It's incredibly sad people are trying to stop this innovation by (falsely) claiming AI data centers are drinking all our water
That AI is taking all the jobs
That AI is just a way for billionaires to get rich
All while the technology gets us closer and closer to cures for all diseases
I don't know which foreign country has convinced half of America that AI is the devil, but hopefully this "AI is drinking the water" trend ends sometime soon
We are on the cusp of greatness
11 Grok Bot tips that will make your setup SO much more powerful:
1. Make your first bot a CEO. Then make all your other bots workers the CEO delegates to. Push all tasks through your CEO agent
2. The best way to create bots for yourself is the brain dump -> reverse prompt method. Brain dump EVERYTHING about yourself then ask your agent what are the best agents you can make that will increase your productivity
3. Reverse prompt your routines. After setting up your bots, ask your CEO agent what are the best routines each bot can set up to increase your productivity.
4. Set up your plugins. My favorite one is AgentMail. It’s free and lets you give your agents an email inbox so they can then sign up for their own accounts instead of using yours.
5. If you have multiple devices and computers, download Tailscale (it's free) on all of your devices. This creates a private network which will allow your agents to move across all your computers and manage them however you want
6. Connection your agent to GitHub/Vercel/Cursor Origin so it can vibe code for you. Will write code and push the code live autonomously for you
7. Give them actual names, not just ‘coder’ or ‘marketer’. Way more fun this way. I just enjoy communicating with agents that have real names so much more. Put their roles in the title instead (will display next to their names).
8. Grok Bot has its own limits outside of Grok Build and Cursor, so don’t feel like you need to limit yourself on the other tools to get the most out of Grok Bot
9. If you have a bunch of different agents, create 'sections' to organize them into folders
10. They by default do all of their work in the virtual machine. if you want them to do work locally on your computers, explicitly ask them to do work on your computer
11. Install the Notion plugin then have your agents record their work in a Notion note. You can then review this later to track everything they do independently.
Boom, you have a super powered team of AI agents working for you
Grok Bot多agent设置清单很具体，但涉及邮件、部署与本机权限时应先做最小授权。
▸ 折叠2条(转推/噪音)
转推2026-08-20 RT @AlexFinn: Absolutely insane. Major breakthrough today
Moderna used AI to design an mRNA cancer vaccine that succeeded in a phase 3 tri…
转推2026-08-19 RT @AlexFinn: 11 Grok Bot tips that will make your setup SO much more powerful:
1. Make your first bot a CEO. Then make all your other bot…
Daniel Vassallo@dvassalloindie203.6K粉 · 2条本期实际内容集中在家庭用车故障，只有窗口前旧推，未进入正文。
Dear @elonmusk please make an 8-seater EV for families who are contributing to improving the total fertility rate of this county. 🙏
引用@dvassalloDon’t buy a @Kia. We have an 8-seat 2026 Carnival hybrid with a defective 12V system. I have to jump start the car every time but Kia’s diagnostics isn’t showing a fault so they can’t do anything even though the car won’t start at the dealership!! Insane … ↗
Don’t buy a @Kia. We have an 8-seat 2026 Carnival hybrid with a defective 12V system. I have to jump start the car every time but Kia’s diagnostics isn’t showing a fault so they can’t do anything even though the car won’t start at the dealership!! Insane levels of incompetence.
Already been to two different dealerships, same outcome. They gave me back a car that won’t start! 🤯
Like @lulumeservey says
"Everything is fake now. Fake content by fake influencers with fake engagement from fake followers, launching fake products with fake testimonials. A Potemkin village on a global scale, a glossy facade propped up by heaps and heaps of beige slop. Temu for content."
Being real on the internet is special!
引用@zachmoskowJust so everyone knows...we got a quote for a “launch video” package from the company everyone uses.
$17K for the video. +$25K and they hand you 50 influencers who will push it, repost it, and flood the comments so the viewcount goes above 500k.
That’s the … ↗
Found this button to disable PWM on iPhone which apparently reduces display flickering
The wording here is super ambigous but apparenly toggling this to green disables PWM? https://t.co/dFkQok5UO2
引用@grokNo confirmed plans yet.
Next iPhones (18 series) will almost certainly keep the iPhone 17's Accessibility toggle to disable PWM for DC-like dimming. Default remains ~480 Hz PWM on OLED.
OLED MacBook Pros (late 2026/early 2027 redesign) switch to tandem … ↗
It's very easy to detect when a GTA 6 leak is real or not
The camera movements
AI videos keep messing up the camera movements and make them too random
引用@Ramon4lanAbastecer o carro, fazer compras dentro de um supermercado, viajar de trem...
O nível de detalhes no supermercado me surpreende demais.
Espero que seja realmente assim o jogo. https://t.co/ZA1HPo16WO↗
Yes it's real and not healthmaxxing
One third of people now have sleep deprivation or chronic sleep disorders, one third!!!
Our globe has been getting hotter and hotter reaching record heat levels every year somewhere
Our houses are built more insulated than ever, they don't breathe anymore
Together that means people sleeping bad, in bedrooms that are too hot, too humid, ans with CO2 levels that are way too high
That's a modern problem not seen before, it wasn't like that 20 years ago!
引用@Jarneoffgrid@levelsio@grok is this real or are we just going crazy with the healthmaxxing? I feel like temperature is important but CO2 is overblown? I always think of the submariners who could just inhale like 40,000 ppm and still perform normal ↗
Most people sleep like this and then wonder why they have sleep problems
76F/24C is way too hot for sleep (should be 18C/64F)
1237 CO2 ppm is way too high (should be 600 max)
76% humidity is way too high (should be 40%)
引用@eigen_moomini found the vitamin.
fuck me no wonder my sleep has been ass for the past 20 years https://t.co/hJDD6DWC0z↗
引用@forgebitzlooks like reddit is almost wiped from chatgpt sources
the query fanout changes had a big impact
and the past couple of days it seems to be almost completely removed from prompt responses
https://t.co/oCGm9M0yPOhttps://t.co/c2GJF2apG7↗
Related @GeorgesStPierre one of the greatest UFC fighters ever doing his first 100kg/225lbs bench press PR
It shows how completely disconnected most people's perception is of strength
If you can do this you're stronger than 99% of people in society!
https://t.co/GRq5uzbEGY
引用@levelsioMy squat is 125 kg/275 lbs/1.6x bw (intermediate)
My bench is 100 kg/220 lbs/1.3x bw (intermediate)
My deadlift is 125 kg/275 lbs/1.6x bw (novice)
I think wanna move slowly from intermediate to advanced I think, I don't need to be elite lifter
My deadlift … ↗
My squat is 125 kg/275 lbs/1.6x bw (intermediate)
My bench is 100 kg/220 lbs/1.3x bw (intermediate)
My deadlift is 125 kg/275 lbs/1.6x bw (novice)
I think wanna move slowly from intermediate to advanced I think, I don't need to be elite lifter
My deadlift now is way too low for how heavy my squat and bench are, but that's because I've always been scared of breaking my back with it, so I was always careful, so I am fixing that now
But I am very very proud of myself, I used to be skinny 50 kg and weak! Now I'm 75kg and very strong!
Intermediate level lifting is stronger than ~50% of lifters (not society!), advanced is stronger than ~80%, elite is stronger than ~95%. If you look society wise, these lifts make me stronger than 95% of people, because most people don't even lift!
I always find it funny when you post lifts you get anon accounts calling you weak, which makes no sense, just going to the gym 3x per week, you're probably stronger than 95% of people! Which for me is a big achievement 👏
引用@Lordcostanza@levelsio What even are you fitness goals. The way you talk about health, I would’ve thought you deadlift more… ↗
🇹🇭 Nomads and expats in Thailand seem to be in the crosshairs of the Thai government and Bank of Thailand this year
Foreigners who have a bank account on a DTV visa are getting them closed, and those trying to open new bank accounts are getting rejected
Also foreigners who own real estate in Thailand through joint ventures with local Thai (a decades long loophole to own real estate where it's actually illegal to own real estate as a foreigner) are now getting forcefully liquidated all over the country
引用@nomadscom🇹🇭 Digital nomads are cracking Thailand’s banking puzzle before October arrivals:
🧳 Bangkok-bound ecommerce founder from Toronto is scouting co-working buddies for October arrival.
🏦 Wise keeps Dutch addresses active for some users, but Thai spending habits … ↗
引用@levelsioToday's lunch:
100g protein
1000 kcal
Featuring:
- complete protein
- heme iron
- zinc
- b12
- potassium
- antioxidants
- anthocyanins
- vitamin c
- fiber
- manganese
And 😋 https://t.co/qWphKRgxbi↗
▸ 折叠3条(转推/噪音)
转推2026-08-20 RT @denisyurchak: I quit my job to work on my own apps. Now my day looks like this:
9:00: wake up with no alarm clock (#1 perk of a nice b…
转推2026-08-19 RT @MaxAnderson: > be me
> take wife to see machu picchu
> stray cat follows us home, wife wants to keep it
> last week of our trip spe…
made a thing for your next brand project.
oddgrid — generative pixel grid art.
pick a look, set your palette, add motion. export as PNG, vector SVG, animated SVG, or video.
free, no signup, runs entirely in your browser. https://t.co/khGw3iaJ25
Oddgrid免费、免注册、浏览器本地运行，还支持PNG、SVG、动画和视频导出。
Marc Lou@marclouindie372.4K粉 · 8条TrustMRR做到150笔收购并长出用户自建subreddit，同时快速发布农残/营养小工具。
Done ✅
Now showing pesticides per country 🌎
The US 🇺🇸 has relatively more residue than Canada 🇨🇦
and European countries 🇫🇷🇩🇪🇺🇸
Some random facts:
🥔 Potatoes have a high load in the US
🥑 Avocado has very low residue in the US https://t.co/eQ1wA7VrqQ
引用@dr@marclou Can you do region only data? Eg I’m in the EU so don’t care about testing/results from the US (totally different produce systems) ↗
I came to Singapore in 2018 for a visa run from Bali.
At the time, I didn’t know any indie hackers, and I felt very lonely.
Little did I know that I’d come back to the same park 8 years later and around 100 indie hackers would show up!
Sorry we had to leave early to catch our flight. Thank you all for coming ❤️
引用@marclouI just landed in Singapore 🇸🇬
Let's meet for the 11th Indie Hacker Meetup!
📍 East Coast Park
📅 Wed, August 19
🕣 7:00–8:00 PM
People have met cofounders, built startups together, and made real friendships at these meetups, so join us ❤️
RSVP: … ↗
I updated TrustMRR’s paywall with a nice wall of love 🥰
It also shows how many startups have been acquired through the marketplace, and for the acquisitions that were made public, the sale price. https://t.co/awt9IPYWDL
I made a mini-app that compares pesticide and longevity for 69 fruits, vegetables, and beans.
It uses 124,880 USDA + UK samples, EPA toxicity benchmarks, and Food Compass 2.0. Click a fruit to see more.
https://t.co/lqryeeYGY2
Take #1: The healthiest fruits/veggies have the most pesticides. Fuck.
Take #2: Beans are superfoods.
引用@levelsioI used to put blueberries in there
Until a few weeks ago
When I discovered that blueberries are some of the most pesticide contaminated things you can eat
They have thin skin and are tiny so they're pesticide bombs
The thicker the fruit the better I … ↗
引用@trust_mrrACQUIRED ✅ 🎉
A SaaS for image generation just got acquired on TrustMRR.
💰 Sold for $5K
🤑 $423 revenue last 30 days
📈 1.0x multiple
It took 35 days from listing. Huge congrats to the seller, the buyer, and the broker @marclou 👏
Acquisition #150 on TrustMRR … ↗
转推2026-08-19 RT @marclou: I updated TrustMRR’s paywall with a nice wall of love 🥰
It also shows how many startups have been acquired through the market…
转推2026-08-19 RT @FiachraRM: So cool to meet @jackfriks parents today @wonjitos & @marclou ! Marc and I also said “moo!” when we took a selfie 🐄 (iykyk)…
Jon Yongfook@yongfookindie168.7K粉 · 5条一边用AI做设计，一边思考营销站该怎样给agent与MCP安排一级入口。
With AI I'm doing some of the best design work of my life, but functionally I have devolved into that annoying micromanage-y creative director type who is like
"nope move that box to the left. not that much. NOPE" https://t.co/QRCaT3x6bP
Idea for SaaS:
AI agent that observes your competitors "feature comparison with <your app>" lists. Builds the features they say you don't have. Then auto files a trade libel dispute requiring them to remove false information.
Comparison lists are dumb in a post AI world.
Where does the "AI" section (info about mcp, chat plugins etc) live on a marketing site. Is it under integrations, developers...?
Or does every site need a first class "AI Agents" section in the nav now.
Discuss.
营销站是否需要一级AI Agents导航，是agent成为正式使用者的早期信息架构问题。
Adam Lyttle@adamlyttleappsindie57.2K粉 · 8条完整复盘AI游戏的20小时失控循环，并改用更窄的视觉精修冲刺。
What I've learned about vibe coding a game (so far):
1.
AI is very good at building the initial prototype game engine. Here you can test how the game mechanics work, whether its' enjoyable and what needs to be tweaked.
At this stage you have a PS1 era game. Horray.
There's no point going any further with a game if it just isn't fun or is buggy at this stage. I think the key is to polish the actual gameplay here as much as possible before diving into the graphics.
2.
It's much easier to just buy 3d game assets from places like TurboSquid than it is to get AI to generate the 3d meshes for you.
Then you use the initial mesh as the base then use the fal ai mcp to make modifications to the 3d mesh and tweak textures. You don't need to learn any 3d modelling you can just talk to your agent using normal language ("make the hull longer")
Now you have an Unreal engine 1 era game. Horray!
3.
I'm still learning the next phase. But clearly the next big jump is refinement. And this is the part that goes way outside my skillset and understanding. So I'm experimenting:
My current thinking is utilising very specific and concentrated sprints with gauntlet loops to improve one part of the game visuals at a time: ambience, lighting, etc.
I've seen others do amazing things with this push.
引用@adamlyttleappsGauntlet loop round 2: lighting + ambience
First image = current game screenshot
Second image = mockup of what I want it to look like
I've mocked up how I want the game to look. Took a scene from the game that shows the desired outcome. I used nano banana … ↗
Gauntlet loop round 2: lighting + ambience
First image = current game screenshot
Second image = mockup of what I want it to look like
I've mocked up how I want the game to look. Took a scene from the game that shows the desired outcome. I used nano banana to mockup just ambience + lighting updates.
Now I've directed sol to make it happen using the gauntlet loop.
Perhaps my last loop was too ambitious. Which was what caused it to keep going off track. Perhaps the key is mini sprints.
This case just update the lighting, ambience, water and sky.
I intentionally didn't mention anything about performance. I don't want to waste loops on micro testing every change. Instead: I'll tackle performance in another gauntlet loop session (if this one even works)
Going to bed. Let's see what it looks like in the morning
He’s out here again with the stall in public
Last time it was a word game now it’s Pokédex for pets
This is what dedication looks like. I need to take note
This passed by without me even realising:
11 years sober this month
引用@adamlyttleappsToday I am 8 years sober
What has changed in my life:
— healthier relationships
— purpose and direction
— attention span
— reliability
— self respect 🫡 ↗
引用@adamlyttleappsVibecoded a 20v20 PvP seafaring combat game with the kids last night
They came up with the ideas and we prompted Claude together
It’s called SEA BARONS:
Captain a ship or ride along
Attack other ships
Claim territory (islands)
Defend your island from the … ↗
I like this take
Did many flash games go on to be turned into successful (monetised) games?
引用@Aidan_WolfAI Three.js games are Flash all over again.
Cheap. Janky. Instant. Fun. Made by one or two people and played by millions. AAA was worried then what was going to happen.
But Flash didn’t kill AAA. It forced the industry to adopt digital distribution and we … ↗
I had high hopes for this.
But I'm abandoning the path it's on
There's this sunk loss falacy.. its been running for 20 hours may as well let it keep going. But the results are just not good.
引用@adamlyttleappsGauntlet loop results so far 🤷🏻♂️
20 hours later and I'm finding the loop goes off track too easily
In the first 4 hours it generated the 3d asset for the ship and the person. GREAT!
But after that got stuck in a loop of its own creation
The problem: … ↗
Gauntlet loop results so far 🤷🏻♂️
20 hours later and I'm finding the loop goes off track too easily
In the first 4 hours it generated the 3d asset for the ship and the person. GREAT!
But after that got stuck in a loop of its own creation
The problem: fal ai mcp was unreachable for a request. So codex in all its wisdom assumed procedural asset generation was good enough and never called fal ai again.
This resulted in poor generated assets that were sent to the subagent for review. Rejected. Then generated again. Rejected. To infinity
The only way I found out: my fal ai balance wasn't being debited.
And the only solution was to run another agent to clean up the mess. And a strict rule that halts the loop if fal ai becomes unreachable with a fallback to retry again later.
The result: not much progress after the initial 4 hours.
16 hours of token burning.
引用@adamlyttleapps🏴☠️ Sea Barons: My free to play multiplayer shoot-em-up has been updated:
- Now 4 teams of 5
- You can invite your friends into your team
- Plus you can now CHAT
There's also been a major overhaul to gameplay based on feedback:
- More arcade like … ↗
I’m quite curious what the future of marketing looks like when all question/answer/exploratory queries are 100% contained to a single platform (Google, ChatGPT, Claude, etc).
引用@Charles_SEOTHIS IS WHAT GOOGLE WILL PROBABLY LOOK LIKE IN 2028 🚨
AI Mode is just the BETA...
What Google is moving towards is a split interface.
Left side ANSWERS you - Right side DOES something.
Sometimes it generates a tool on the fly, e.g. a calculator, a … ↗
the https://t.co/wy8pUhuFoS beta of our iOS app has been a huge success. it's become the primary way i use the tool. just feels good.
submitted for proper public release! 🤞
curious to see how a $129/year in-app purchase lands 🤪 https://t.co/92gAYC99xk
i'm *incredibly* proud of the software i've been building this year. especially over the past few months i feel like i've really found my footing on tooling & processes to create actually great software.
maybe nobody else notices or cares, but just really proud of myself.
in a previous life I was designing the campaign website for a presidential candidate and the feedback was “make the blue more presidential”. which could not have been more annoying to 25 y/o Josh.
I now spend my entire day just telling LLMs to make things more presidential.
引用@yongfookWith AI I'm doing some of the best design work of my life, but functionally I have devolved into that annoying micromanage-y creative director type who is like
"nope move that box to the left. not that much. NOPE" https://t.co/QRCaT3x6bP↗
first https://t.co/rHttZyJqV0 takedown succeeded!
got a shopify store? your products, images and brand are almost certainly being stolen/ripped off.
let knockoff handle all of the riffraff for you. 🤗 https://t.co/DbZhOzvnyt
引用@forgebitzlooks like reddit is almost wiped from chatgpt sources
the query fanout changes had a big impact
and the past couple of days it seems to be almost completely removed from prompt responses
https://t.co/oCGm9M0yPOhttps://t.co/c2GJF2apG7↗
just ran a full audit on the @maybe figma account and there are over FIVE THOUSAND screens from over 5 years of work.
every single one designed by @justinmfarrugia. absolute beast of a designer.
引用@Shpigfordif anyone wants to purchase thousands of detailed Figma design assets as we're shutting @maybe down...DM me. 🙂
would probably be pretty valuable for anyone building models for design. ↗
DigitalOcean MCP + Hatchbox MCP + Cloudflare MCP is absolute heaven for someone like me who's been begrudgingly doing sysadmin junk on his projects for 25 years.
Pure sorcery.
That feeling when you run a podcast business and your favorite living philosopher (who you read zealously at uni) sends you a casual email about becoming a customer with their pod.
Podscan has a lot of well-known customers. But I really have to hold back the inner fanboy today.
The fact that a major part of mitigation was 403ing Copilot requests says it all. Even “just” smart autocorrect now operates at a scale that breaks the largest players when a tiny thing isn’t perfect.
Feels like the internet is getting even more brittle with agents & AI tools.
引用@AdamRackisThe jokes and memes are funny just so long as we all realize GitLab, BitBucket or any of the others would have keeled over and died at half the volume of ai slop GitHub has been deluged with 😂 ↗
AI请求规模让GitHub一类基础设施更脆；窗口前旧推，仅留附录。
Tony Dinh@tdinh_meindie198.1K粉 · 2条用Claude做小游戏做得很开心，也现场撞上NaNK这类原型期类型错误。
ok Claude is in trouble
vibe coding is fun until you see damage number shows up as "NaNK" 🤣
Should have asked Claude to use TypeScript https://t.co/V66InKBma0
first creator payout sent 🎉
experimenting with my new SaaS:
creators <> brands
finding brands is easy but finding high-quality video creators is hard
if you're looking to earn money through sponsorships, reply to this and we might talk 💸 https://t.co/q1L1bDY4gY
biggest takeaway:
Stripe billing is growing 70%+ Y/Y
Stripe is the internet economy and it's booming
and AI impact just getting started https://t.co/w5TajU0Sgb
引用@EricNewcomerHere's Stripe's letter to investors explaining its acquisition of OpenRouter (LEAKED) https://t.co/IDH5rIbDq0↗
last month, a lot of strangers made me $30,000
it's very simple:
- I have a product that works (like Outrank)
- they have influence over buyers
- they work as affiliates and get me customers
but how do you find affiliates to promote your products?
answering that in the newsletter tomorrow
subscribe to get the email 👇
signups for the tmaker Founders Room are open for 48 hours
being upfront about this: yes, Marc Lou is doing an AMA with members on Tuesday. a lot of people asked to join because of it
don't join for that
a guest is in the room for 1-2 hours. the real value is the other 150+ bootstrapped founders solving each other's problems in a room small enough that people know your name
if Marc is the only reason you want in, save your $99
if you want the room, it's open until Friday
👉 https://t.co/uh67w9ry0Q
then closed again
the product is so good that we cannot serve everyone and we had to close sign ups
hope to accept new people early next week
引用@ArnaudBelingaCXI tested 20+ Reddit tools.
I always got disappointed.
Then I tried Bazzly, loved it and got features I wanted within 2 hours by asking the support ⚡
I shared it in a lead magnet, nobody cared.
Meanwhile I get clients from Reddit every week because of it. ↗
▸ 折叠2条(转推/噪音)
转推2026-08-20 RT @virgilerietsch: You should join Tibo’s creator platform.
I got paid 50$ for a video I recorded in 10 minutes.
It ain’t much but it co…
转推2026-08-19 RT @nicolasmore_: just joined TMAKER from @tibo_maker
I've been following Tibo for a while and really admire how insanely good he is at d…
Simon Høiberg@SimonHoibergindie161.2K粉 · 6条集中推广Grok Bot与自己的赞助位，也分享本地Qwen驱动的Agent OS体验。
The people claiming we're all in a simulation and someone is watching us this way right now - it gets a little bit easier to believe every day these days.
引用@SimonHoiberg- Cool project goes live.
- I move fast and sponsor it with https://t.co/7NOLCg7lFb.
- Elon Musk shares it.
- The project goes viral.
- I'm making 💸
Being an online entrepreneur is dope 🤩 ↗
I changed my custom Agent OS (OpsLayer) to run on qwen3.5 27B (Q8). I'm MINDBLOWN.
I can't believe how good this is.
Sovereign intelligence is finally here.
引用@SimonHoibergI built something very similar 😁 I call it OpsLayer.
- Agentic Team OS.
- Chat/project board.
- Files and knowledge base.
- Secure API calls with approval.
- Secrets management (AI can't see keys).
- Eval and guardrail management.
- Desktop client and … ↗
- Cool project goes live.
- I move fast and sponsor it with https://t.co/7NOLCg7lFb.
- Elon Musk shares it.
- The project goes viral.
- I'm making 💸
Being an online entrepreneur is dope 🤩
引用@elonmuskShare your Grok @Bot design with others! ↗
转推2026-08-20 RT @petecallaghans: Grok Bot is how I stopped being the person who has to remember everything. Here's how I set it up for my SaaS. https://…
转推2026-08-20 RT @jonathan_wilke: This is crazy! https://t.co/ZElVA2hguv hit 3000 visitors in less than 12 hours with the top bids already crossing $200…
Peter Askew@searchboundindie26.1K粉 · 1条只有一条关于私密点赞和meme账号的轻松发言，本期无业务信号。
Frow now on, users who reach a streak of 365 days have a lifetime 50% discount on Uneed pro, our subscription 🥳
As a reminder, here's what you get with it:
📚 Unlimited product lists
🔁 1 free relaunch per month
🔒 Access to our private discord
📣 Unlimited testimonials and widgets
📈 Domain rating tracking with email alerts
🤑 $1600 on partners perks: discounts, offers etc
Dan Kulkov@DanKulkovindie48.9K粉 · 5条继续小步发布ActivationPal，分享技能仓库星数、内容增长和freemium定价情绪。
If you have 0 customers you don't need:
> an audience
> SEO
> a viral launch video
> paid ads
> content marketing
> growth hacks
You need to find 10 people who have the problem and show them what you built.
DM them. Email them. Message them on LinkedIn.
Just start talking.
If I was at zero users with no audience this is what I'd do:
1. Go through my network and find anyone who has the problem
2. Message them and have a normal convo
3. Show them what I'm building
4. Ask if it would be useful
If they say no:
Do you know anyone who would benefit from this? Can you introduce me?
If no-one responds to your cold outreach you're sending the wrong offer to the wrong people.
You should be getting at least some replies.
Here's an tactic that worked for me for my ticketing software:
- Emailed event organisers (highly targeted)
- Acknowledged their current problem (high fees + slow payouts)
- Led with a concrete offer: 2% + 60p per ticket, paid every 7 days
- Made it stupid easy to reply
I sent them a clear offer solving an obvious financial pain point.
You don't need an audience it's BS.
I got 20k customers for my first product with zero audience.
✅ Cold emails
✅ Email marketing
✅ SEO
An audience is useful for launching stuff and testing the market but it's not a long-term strategy.
It can definitely help I'm not going to deny that but it won't just magically bring new customers every day the way people believe.
95% of stuff I've launched over the years failed
I don't feel embarrassed about it and I usually share why it failed
But sometimes I worry it's not good for my reputation
Then again, I'd never trust anyone who only posts about successes
Better to try lots of stuff and openly fail or no?
The Premier League starts tomorrow.
I built a cool email that mimics old newspaper roundups.
Yesterday's scores, the table, the scorers.
In your inbox. Always free.
Sign up to get the first edition on Saturday morning!
https://t.co/iSlXrzqTtxhttps://t.co/Zgn17dS3GD