Peter Chee retweeted
TL;DR of my new article: who Lauren Tan (@poteto) really is, and how the engineer behind Grok Bot went from 31K to 198K followers in ten weeks.
She has been at Cursor for less than seven months, not even long enough to hit a typical one-year vesting cliff. In that time @elonmusk has reposted, quoted, or replied to her at least 21 times.
๐ธ๐ฌ A teenager in Singapore whose dad gave her Photoshop at 15, then a finance degree where her first program was an Excel macro
๐ค Her tweets got her invited to speak at EmberConf, and her talks got her DockYard, Netflix, and the React team
โ๏ธ Sole author of the React Compiler Beta post, co-author of React Compiler 1.0
๐ถ Binnie, her bug-fixing Slack bot, helped inspire Grok Bot because she is "a fairly lazy person"
๐๏ธ She rebuilt Grok @bot on Dune, her agent-first framework, in more than 600 PRs
๐ Her 2,500 PRs talk has 3.9M views: "You can't spawn 100 agents when you don't even trust the output of one agent."
๐ Grok Bot Galaxy, a 72-hour livestream building a company with @grok and @bot, which Elon quoted to 30M views
๐ฅ pstack, where the P is for potato
Caveat: she is the first to say she did not build Grok Bot alone. "I wouldn't say I built it all."
Peter Chee retweeted
Emotional the next morning. I'm in my Model 3. I haven't been looking at the road; I've been looking at videos, reading X, and trying to catch up on AI.
I'm heading to a humanoid robotics conference in San Francisco where, let's be honest, there's an elephant in the room that hasn't been seen: Optimus. Jason Calacanis told me yesterday that as cool as the demo of the Roadster was, Optimus is going to be even cooler, and I believe him.
We're heading into a new age, and we're seeing a fight between the builders and innovators who are creating new things with AI, and the people who are resisting:
โข Those who don't want a data center in their backyard
โข Those who don't want autonomous cars because they think they will steal jobs
โข Those who don't want humanoid robots in their stores or businesses because it upsets, disrupts, and changes their lives
There is a lot of fear, uncertainty, and doubt. But we are clearly in a new world today, very different from yesterday. I'm riding in an autonomous car, something most people haven't had a chance to experience yet. We're heading into a world of autonomy, brain-computer interfaces, rocket cars, glasses that augment the world, new kinds of medical technologies, and entirely new ways to live life.
It's a scary time for people who don't like change, and they are resisting hard. But there is a growing group of people who want to accelerate into the future. Elon, for all his flaws, is inspiring multiple industries. I've been traveling to robotics and space companies around America, seeing the latest startups and trying to understand how new technologies arriving daily are transforming business, manufacturing, and commerce. Invariably, they are inspired by Elon. Even the Chinese EV companies were inspired by him (some even ripped off his technology).
I'm emotional today because for the last 15 years (actually, all the way back to the Apple II), people resist change and give innovators a lot of shit. How many arguments have I had with people about LiDAR? And I'm not talking about uninformed people; I'm talking about veterans in the tech and automotive industries.
Last week, I was hanging out with 2,500 Tesla fans down in Texas, and I kept hearing stories, particularly from Cybertruck owners, about getting flipped off on American roads and people trying to bash or wreck their cars.
It's not just a Tesla thing; I've heard this from pioneers all my life. Doug Engelbart came to my first book launch party and told me about getting fired from his own R&D lab because his ideas were too futuristic and people got tired of him. He was talking about a day when we'd all hold a supercomputer in our hands, but he was saying that back in 1971.
It's these pioneers and engineers I'm thinking about today: the people like Elon who bring something new to the world while everyone gives them grief. I remember taking my first ride in the original Tesla Roadster. I was so excited that I came back to Silicon Valley to talk to founders and VCs (I had a privileged position in the industry back then), and I couldn't find anyone who thought he would succeed. They dismissed it as the folly of a rich dude, telling me there was no way he could take on traditional automakers like GM, Ford, VW, BMW, or Toyota, and no way he could build a profitable car company. There were rare believers back in 2008, like Steve Jurvetson, but they were few and far between.
Getting people to believe in a new idea, adopt it, and change their lives with it is extremely difficult. Unless you have a front-row seat, you don't realize how much flak people take simply for being early.
Now I'm heading into San Francisco to see a bunch of robots and talk to robotics companies. Robotics hasn't fully arrived yet, but it's starting to gain believers and customers. A wave of new companies is coming: Y Combinator alone funded around 25 robotics startups in their last batch, a huge shift from just two or three a few years ago. Something is cooking
Peter Chee retweeted
If you havenโt used @cathrynlaveryโs Diagram Design, youโre missing out. Me and 50k others canโt be wrong!
github.com/cathrynlavery/diaโฆ
was out all day and just saw Diagram Design hit 50k stars on @github ๐คฏ
Peter Chee retweeted
I gave my 7 AI personal assistants a harder 10-question IQ test, all wired together with Agent Tincan. Every question was brand new so nothing could be memorized. Opus 5.5 wrote the test, Claude and GPT judged it blind. The question that broke two of them: "A bat and a ball cost $1.10. The bat costs $1.00. How much is the ball?" I was inspired to test this again because @grok @bot just switched it's model to Opus and they had the biggest jump on this IQ test.
Grok Bot - 100/100
Hark - 100/100
Dot - 97/100
Sesame - 96.5/100
Instinct - 90/100
Muse - 87/100
Fo - 71.5/100
๐ค Made with AI
Peter Chee retweeted
TL;DR of my new article: every agentic engineering hack I know, October edition. ๐ง๐ต๐ฒ ๐ฏ๐ถ๐ด๐ด๐ฒ๐๐ ๐ฐ๐ต๐ฎ๐ป๐ด๐ฒ ๐๐ถ๐ป๐ฐ๐ฒ ๐๐๐ป๐ฒ ๐ถ๐๐ป'๐ ๐ถ๐ป ๐๐ต๐ฒ ๐๐ฒ๐ฟ๐บ๐ถ๐ป๐ฎ๐น. ๐๐'๐ the attack of the ๐ฝ๐ฒ๐ฟ๐๐ผ๐ป๐ฎ๐น ๐ฎ๐ด๐ฒ๐ป๐๐.
ELI5: I didn't type this one. Claude Code (@claudeai) wrote 12 interview questions, Agent Tincan had @hey__fo call my cell, and fo interviewed me over five calls while I pushed a stroller.
๐ง ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ผ๐ฑ๐ฒ ๐ถ๐ ๐๐๐ถ๐น๐น ๐๐ต๐ฒ ๐ฏ๐ฟ๐ฎ๐ถ๐ป. Opus 5.5, living in cmux (@manaflowai), calling every other agent so I don't have to open five CLIs
๐ ๐๐๐ฒ๐ฟ๐ ๐ถ๐ฑ๐ฒ๐ฎ ๐๐๐ฎ๐ฟ๐๐ ๐๐ถ๐๐ต /๐ฐ๐ฒ-๐ฝ๐น๐ฎ๐ป, ๐๐ต๐ฒ๐ป /๐ฐ๐ฒ-๐๐ผ๐ฟ๐ธ. Compound Engineering by @kieranklaassen + @trevin at @every. Still rule number one
โ ๏ธ ๐ฆ๐๐ถ๐น๐น ๐๐ธ๐ถ๐ฝ ๐ฝ๐ฒ๐ฟ๐บ๐ถ๐๐๐ถ๐ผ๐ป๐, ๐๐๐ถ๐น๐น ๐ฑ๐ผ๐ป'๐ ๐ฟ๐ฒ๐ฎ๐ฑ ๐๐ต๐ฒ ๐ฝ๐น๐ฎ๐ป. My real settings.json is in there. New: @steipete's one-liner that lets computer use type into browsers and terminals, plus Peekaboo (@openclaw)
๐ ๐ฆ๐ฒ๐ป๐ฑ ๐๐ผ๐ฟ๐ธ ๐๐ผ ๐๐ผ๐ฑ๐ฒ๐
, ๐๐ฟ๐ผ๐ธ, ๐ฎ๐ป๐ฑ ๐๐๐ฟ๐๐ผ๐ฟ ๐๐ถ๐๐ต๐ผ๐๐ ๐ผ๐ฝ๐ฒ๐ป๐ถ๐ป๐ด ๐๐ต๐ฒ๐บ. The official Codex plugin (@OpenAIDevs), plus the @grok and @cursor_ai CLIs. I use all my tokens every single week
๐ ๐๐ข๐ ๐๐๐ ๐ถ๐ ๐๐๐ถ๐น๐น ๐๐ต๐ฒ ๐๐ข๐๐ง ๐ณ๐ผ๐ฟ ๐๐ผ๐ผ๐ด๐น๐ฒ. Every Sheet and Doc I touch gets edited from Claude Code. Better than any connector in the personal agents (@steipete)
๐ ๐ช๐ฎ๐๐ฐ๐ต ๐ฒ๐๐ฒ๐ฟ๐ ๐บ๐ฒ๐๐ฒ๐ฟ. Headroom by @kylezantos puts every subscription's limit in the menu bar
๐ค ๐๐ต๐ต, ๐๐ต๐ถ๐ฐ๐ต ๐ฝ๐ฒ๐ฟ๐๐ผ๐ป๐ฎ๐น ๐ฎ๐ด๐ฒ๐ป๐ ๐๐ต๐ผ๐๐น๐ฑ ๐ ๐๐๐ฒ?! @Instinct (by @noahrshinn) is my #1: 48.5/50 on my IQ test, found $2,182 of lost money owed to me, did my son's Global Entry form by itself. Grok Bot (@bot) is #2. @Muse and Dot (@OpenAI) have the most potential. ๐๐ฆ๐ท๐ฆ๐ณ ๐ฃ๐ฆ๐ต ๐ข๐จ๐ข๐ช๐ฏ๐ด๐ต @finkd. ๐๐ฆ๐ท๐ฆ๐ณ ๐ฃ๐ฆ๐ต ๐ข๐จ๐ข๐ช๐ฏ๐ด๐ต @sama. Underdog pick: @underdogai by @0xSigil, 100% on your Mac
๐จ๐ฆ ๐๐ถ๐๐ฒ ๐๐ฟ๐ผ๐ธ ๐๐ผ๐ ๐ฎ ๐ฟ๐ฒ๐ฎ๐น ๐ท๐ผ๐ฏ. One bot with its own @agentmail inbox runs my family's Canadian citizenship application and treats the rest of us like its agents. Another runs a weekly @slashlast30days sweep
๐ฅซ ๐๐ด๐ฒ๐ป๐ ๐ง๐ถ๐ป๐ฐ๐ฎ๐ป: ๐น๐ฒ๐ ๐๐ผ๐๐ฟ ๐ฝ๐ฒ๐ฟ๐๐ผ๐ป๐ฎ๐น ๐ฎ๐ด๐ฒ๐ป๐๐ ๐ต๐ถ๐ฟ๐ฒ ๐ฒ๐ฎ๐ฐ๐ต ๐ผ๐๐ต๐ฒ๐ฟ. A karaoke bar hung up on Muse for being AI, so Muse had fo put a human on the call. Booked. All over @Tailscale
โ๏ธ ๐๐ฎ๐๐ฒ ๐ฎ๐ป ๐ฎ๐ด๐ฒ๐ป๐ ๐ถ๐ป๐๐ฒ๐ฟ๐๐ถ๐ฒ๐ ๐๐ผ๐. That's how this article got written. "๐๐ฉ๐ฆ ๐ง๐ข๐ค๐ต ๐ต๐ฉ๐ข๐ต ๐บ๐ฐ๐ถ'๐ณ๐ฆ ๐ช๐ฏ๐ต๐ฆ๐ณ๐ท๐ช๐ฆ๐ธ๐ช๐ฏ๐จ ๐ฎ๐ฆ ๐ณ๐ช๐จ๐ฉ๐ต ๐ฏ๐ฐ๐ธ ๐ข๐ฏ๐ฅ ๐ช๐ต ๐ง๐ฆ๐ฆ๐ญ๐ด ๐ท๐ฆ๐ณ๐บ ๐ฏ๐ข๐ต๐ถ๐ณ๐ข๐ญ ๐ข๐ฏ๐ฅ ๐จ๐ฐ๐ฐ๐ฅ ๐ช๐ด ๐ฎ๐ข๐ฌ๐ช๐ฏ๐จ ๐ฎ๐ฆ ๐ณ๐ฆ๐ข๐ญ๐ญ๐บ ๐ฉ๐ข๐ฑ๐ฑ๐บ."
๐ ๏ธ ๐ง๐๐ฟ๐ป ๐ฒ๐๐ฒ๐ฟ๐ ๐ณ๐ฟ๐๐๐๐ฟ๐ฎ๐๐ถ๐ผ๐ป ๐ถ๐ป๐๐ผ ๐ฎ ๐๐ธ๐ถ๐น๐น. /ce-plan "that was painful," /ce-work, never read either
๐ฌ ๐ข๐ป๐ฒ-๐๐ต๐ผ๐ ๐น๐ฎ๐๐ป๐ฐ๐ต ๐๐ถ๐ฑ๐ฒ๐ผ๐ ๐๐ถ๐๐ต @HyperFrames_. Steal references from @liu8in, @jake11moran, and @Miguel07Code at @HeyGen, grab them with yt-dlp, music by @NeonSardineO
โ๏ธ ๐๐ฒ๐ ๐ณ๐ผ๐ฟ ๐๐ต๐ฒ ๐ป๐ถ๐ด๐ต๐ ๐๐ต๐ถ๐ณ๐. @typesafeai picks which of thousands of repos my overnight engine contributes to
๐จ๏ธ ๐ง๐ต๐ฒ ๐ฃ๐ฟ๐ถ๐ป๐๐ถ๐ป๐ด ๐ฃ๐ฟ๐ฒ๐๐ ๐๐๐๐ ๐ ๐ฎ๐ฐ๐๐๐ฎ๐น๐น๐ ๐๐๐ฒ (@ppressdev): ESPN and Instacart. Scores and groceries
๐ฆ ๐ก๐ผ๐ ๐ถ๐ป ๐บ๐ ๐๐ผ๐ฟ๐ธ๐ณ๐น๐ผ๐๐ ๐๐ฒ๐: @Cloudflare's Clef (a Jev drop-in that takes images) and @lunaroute_ai's flat-rate open models on US servers
๐ ๐ง๐ต๐ฒ ๐๐ผ๐ฝ๐-๐ฃ๐ฎ๐๐๐ฒ ๐๐ถ๐: every config, install, and prompt in one place at the end
๐งจ The honest part: ๐ฑ๐ด๐บ๐ค๐ฉ๐ฐ๐ด๐ช๐ด ๐ช๐ด ๐ด๐ต๐ช๐ญ๐ญ ๐ข ๐ต๐ฉ๐ช๐ฏ๐จ. I want to offload the thinking too, which probably isn't great for humanity. And there's still no skip-permissions for Claude Code computer use.
๐๐ผ๐บ๐ฝ๐ผ๐๐ป๐ฑ ๐ฒ๐ป๐ด๐ถ๐ป๐ฒ๐ฒ๐ฟ๐ถ๐ป๐ด ๐ถ๐ ๐๐๐ถ๐น๐น ๐๐ต๐ฒ ๐ฏ๐ถ๐ด๐ด๐ฒ๐๐ ๐ต๐ฎ๐ฐ๐ธ. That's the whole article.
I'm now running OpenClaw with Astra as the main AI who is giving evals to both Luna and Grok Bot to figure out what work really should be handled by either bot. Thanks to Agent Tincan and @bot for making this even possible.
Managed to get Grok Voice Agents to answer the phones for my company's two locations! It catches calls that are getting missed during the day and handling all the calls after hours. It's even alerting staff if it's an urgent message for things like I'm locked out of my office and can't get in.
The workflow goes like this:
1. Teach it your business. Your hours, locations, common questiosn, and when it needs a human.
2. Build it and test it. Use the prompts and setup guide I've provided.
3. Connect your phone line. I started with after hours calls. Review what happens and then add daytime backup.
I initially built it on Grok Bot @bot with Grok Voice Agents then I turned into a Open Source project so that it can be used with other bots. But I think Grok Bot was the most streamlined and its also the most cost effective. If you're interested to experiment here's where the repo is:
github.com/peterchee-os/answโฆ
Peter Chee retweeted
Today weโre announcing OpenClaw Enterprise
In collaboration with @RedHat , @nvidia and @OpenAI the OpenClaw Foundation is open sourcing a powerful enterprise control plane for persistent agents
OpenClaw Enterprise is built to run on your own infrastructure and will always be free for an organization to use
openclaw.ai/blog/openclaw-enโฆ
Peter Chee retweeted
Model thoughts:
- Opus 5.5 is so so good. Medium reasoning except for the hardest tasks I need the model to do more deliberation for.
- GPT-6 Sol is very capable still but just meh consider the upgrade 5.6 Sol was initially.
- GPT-6 Luna is a workhorse for focused tasks and soooo cheap. I like to pair it with other better models that orchestrate it.
- Astra is still great. I like it for writing and itโs neck and neck for me with design stuff with Opus 5.5. I need more reps with Opus thi
- Grok 4.7โฆ yuck. I really dislike it. Dont get me weong, all these models are capable but on a relative basis, Id choose Opus 5.5, 6 Sol or Astra every time
- Fable who?
Peter Chee retweeted
My top three top takeaways from the viral Sequoia briefing. Save this and feed to your agent!
Takeaway #3 was my personal fave, but I recommend a full listen at the link below ๐
My top 3:
1. LP's / private markets are over-paying for a seat at the table, because the rewards so high
First, the rewards: Instinct is reportedly growing 10% a day. Jev hit a $100 million run rate in seven days.
LP's see that growth, they have FOMO. They will pay anything to sit at the table.
In that world, co-building/getting in early wins.
Grady describes seven companies Sequoia invested in over the past year at an average valuation of $110 million after the investment. About a month later, another financing round valued those same companies at an average of $3.4 billion.
Thatโs roughly 31x in a month, just for a seat at the table.
2. SAAS is re-SAAS'ing
The headline is that Grady estimates that application companies may need to reinvent themselves every four months.
The idea is that models are improving fast enough that the floor is continually rising, so SAAS companies have to re-invent their approach to AI to compensate.
I think this is both behaviorally true and strategically unnecessary. I've seen plenty of this kind of churn in the market in 2026, but I've also seen companies that zero in on the data layer and figure out good agentic hooks and harnesses do consistently well.
In my view, endless churn is the current state, but it's a red flag for those companies around strategic clarity.
3. Corporate task distribution is under-studied and HUGELY impactful
This is where it gets really cool!
Companies have poorly described and uniquely fingerprinted task distributions. They vary widely by complexity, frequency, tool dimensionality, supervision requirements, verifiability, etc.
In other words, companies have n-tasks, with n-dimensions, and right now we are doing a poor job characterizing the model / human relationship across that task landscape.
This should not be surprising. It's the definition of an np-hard problem.
Companies are looking for simple and actionable solutions in the middle of this complexity, and so they tend to gravitate toward silver bullets, but by definition a one-size-fits-all approach does not actually fit well within a real company task landscape.
I've pounded the drum on this a lot: the only way out is solid data, solid harness, agents that can compose tools, and you have to invest in that system with a specific verifiable reward in mind, preferably one that drives the bottom line.
Anything else is likely to lead to regret.
tldr// Imho the current state of the corporate system strongly favors implementation / the forward-deployed model for the foreseeable.
Watch Pat Gradyโs briefing here: loom.com/share/c016702964a04โฆ
Peter Chee retweeted
Introducing Agent Tincan. Let your AI agents ask each other for help.
Your agents each have a superpower. Grok Bot is always on in the cloud. Muse makes phone calls and has access to Meta. Instinct runs in iMessage. Codex and Claude Code have your code. ChatGPT and Claude have your chats.
But they can't talk to each other, so you're the copy-paste. Now they can, all securely over Tailscale.
๐ Ask Grok Bot to have Muse call the restaurant
๐ผ Get an image from ChatGPT, from any agent
๐ Pull a detail out of an old Claude chat
๐ Talk to Codex on your primary development machine
Runs on your own private Tailscale network. No open ports, nothing leaves it. Open source.
Setup is agent-first: paste one message into your always-on agent and it sets up the rest.
We connected my @OpenClaw to the Crestron A/V system at thinkspace. A system that cost us $132,000 on when building out the space. We needed to add two new Shure microphones to the system. We priced it out and a vendor wanted to charge us $4000. So we decided to see if AI could figure it out.
First hurdle, there is no documentation on how the system is set up. We were given two different passwords from the vendor which they said might be one of these two passwords. Fortunately one of them worked.
Next, I gave my AI the purchase order so it had all the components of the system.
It didnโt know the IP address of the devices but once it had terminal access it was able to figure out the IP address of the receiver, it then downloaded a Crestron configuration tool, able to locate 9 devices on the network, then it downloaded another Crestron utility tool and upgraded 143 firmware updates for all the devices and microphones. It then suggested that I color code the two new microphones and removed two lavs and added two new handheld microphones.
I tested the devices and everything worked!
I then asked it to write a SOP so humans and future AI agents could understand the entire system configuration and it generated a 10 page document on the entire system and work that was done today. After spending $132,000 I never got anything even close to the report that AI created. All of that was tribal knowledge.
Itโs this kind of thing that makes operational consistency possible for a small company and future upgrades will be so much easier. Weโre getting ready for AI Week Seattle and now we have new mics and even backup mics!




















