Sean

@yacineMTB speed is natures patent

patents are the most idiotic thing maybe ever

@mcuban it will help with the push toward rent seeking too. the moat between SOTA models and local inference will always be ~6 months due to the control of supply chains. large model providers will always have that edge.

We should federally tax Tokens at the Provider level. Not a lot. Less than 50c per million tokens. It will accomplish 4 things (at least ) 1. It will push the big AI players to optimize tokenization, caching , routing and localization Which will 2. Reduce energy usage. Saving them in energy costs more than what they paid in tax and reducing strain created by the growth in energy consumption Which will 3. Generate maybe 10 billion dollars a year to start, but over the next ten years could grow 30x to 100x Which will 4. Create a source of funding to pay down the federal debt or deploy, in response to the things AI brings that we don’t expect or don’t like At some point the models will pass it on to customers. Of course. That’s ok. Customers will have the ability to choose between providers. Or to do everything using open source models locally. Thoughts ?

@steipete lmao “extremely lean” w/ $1.8m in tokens.

People freaking out over my AI spend. What nobody sees: Part of what excites me so much about working on OpenClaw is that I'm trying to answer the question: How would we build software in the future if tokens don't matter? We constant run ~100 codex in the cloud, reviewing every PR, every issue. If a fix on main lands, @clawsweeper will eventually find that 6 month old issue and close it with an exact reference. We run codex on every commit to review for security issues (as it's far too easy to miss). We run codex to de-duplicate issues and find clusters and send reports for the most pressing issues. We have agents that can recreate complex setups, spin up ephemeral crabbox.sh machines, log into e.g. Telegram, make a video and post before/after fix on the PR. There's codex that watch new issues and - if it fits our documented vision well, automatically create a PR of it. (that then another codex reviews) We have codex running that scans comments for spam and blocks people. We have codex instances running that verify performance benchmarks and report regressions into Discord. We have agents that listen on our meetings and proactively start work, e.g. create PRs when we discuss new features while we discuss them. We build clawpatch.ai to split all our projects into functional units to review and find bugs and regresssions. We do the same split for security with Vercel's deepsec and Codex Security to find regressions and vulnerabilities. All that automation allows us to run this project extremely lean.

@reach_vb need to connect codex to my printer...

Codex literally debugged and fixed the printer! The printer wasn't working via the normal PDF to print route, ask codex to iterate and gave it feedback if it worked or not in 3 turns we had fixed the biggest problem of our generation. Scores a resounding 100% on PrinterBench!

as i continue experimenting with different agent setups there is an inverse correlation to lowered productivity and more agents. having an agent for one small task the is managed by a parent agent just doesn't work ~60% of the time. that 40% time spend fixing it ends up costing more time then just asking codex to do it with a general set of knowledge. the best setup i've found so far are slop bots: N agents work towards goal, each with a different edge, then me or agent validates the best one to move forward with.

@elonmusk @Variety at least it wasn't The Iliad.

@Variety Chris Nolan is pissing on Homer’s grave. Disgraceful.

Elon Musk is bashing Christopher Nolan for #TheOdyssey casting, alleging the director cast Lupita Nyong’o as Helen of Troy because he “wants the awards.” In additional X posts, Musk re-tweeted posts that mocked Elliot Page’s masculinity and claimed that Nolan was stomping all over Homer’s grave because of his casting choices. variety.com/2026/film/news…

@zachlieberman peaceful to the eyes.

Printing again

@NickADobos u can make a demo with live data in the same amount of time as a power point deck — that'll make it that much easier to find the right idea to move forward with.

Haven’t seen many people talking about orchestrating ai agents + human teams at scale Seems like the new meta is: - a bunch of slop cannons & product designers churning out prototypes and ideas - ai reviewing PRs like crazy - light human review Every few weeks have someone do a big refactor pass, make one plan .md doc and have all humans review the plan. Throw out lots. Consolidate slop into full maintainable systems. Big change from previous eras where refactoring was expensive & time consuming. You’ve now got to be much quicker to delete, iterate & rearchitect.

@abcampbell its charts like this that remind you were still animals. tell an ape to challenge itself for growth vs sit in front of a feeder tube of stimulants.

what’s causing this? how do we fix it?

@NickADobos afraid ~2 months and codex is next.

Everyone with an ounce of sense knew this was coming. Did y’all seriously expect to run your insane custom scripts on top of Claude at the massively discounted subscription rate forever?

@skooookum codex’s limits via the subscription are insane. i blew through ~$100 in token via anthropic api yesterday in 2 hours. the similar task done via codex dropped my weekly limit ~5%

Genuinely have no idea what’s happening here. I have >5 agents running in codex 8-10 hours a day and rarely hit my weekly limits.

there’s more room for lists to become a better experience. instead of being only people you add, they can be similar to the new reply options: accounts added to a list + accounts they engage with / follow. would help open the door for finding new accounts producing high quality content in your niche.

building has become more streamlined but broadcasting your idea is now the real moat — how big is your audience and how much do they trust you? you can build an large audience that trusts you with the same speed.

@shannholmberg why have the separate containers instead of running profiles/agents in one env?

my entire Hermes Agent setup is controlled from one folder on a VPS I can manage it from my laptop or phone in under 10 seconds, spin up isolated agents per project, and never lose context here's the full setup: > bash command "ssh hermes" auto-connects to the VPS > session pinned in CMUX terminal for instant access > Claude Code session inside the ssh connection with full remote access > everything runs from a control station folder folder structure: vps-agents/ ├── README .md overview + how to use it ├── CLAUDE .md auto-loaded project context for Claude Code ├── agents/hermes-main/ │ ├── inventory .md IP, hostname, where it runs, who owns it │ ├── docker. md draft docker-compose + first-boot checklist │ ├── runbook. md "it broke, what now" — 7 common scenarios │ ├── env-map. md table of which keys live where (no values) │ └── backup. md GitHub backup flow + restore steps └── shared/ ├── security. md VPS-wide hardening checklist └── commands. md Docker / Hermes / UFW cheat sheet from this session I can spin up new docker-isolated agents per area and each one can have sub-profiles the whole stack is documented, backed up to GitHub, and accessible from anywhere

@codyaims and that’s only going to grow [lower us workforce and higher global %]

China’s manufacturing workforce is ~130M, producing 28% of global mfg at $4.66T USA’s manufacturing workforce is ~12.5M, producing 17% of global mfg at $2.91T America has 10% their workforce producing 62% their value

@perrymetzger it really is just the simple next step to scaling. without these orders of magnitude in productivity, how else could we scale ?

All of the top quality programmers I know who are using AI are suddenly jet powered, operating at 10 or 20 or 30 times the speed that they used to. Projects that used to seem like they required a long time and serious effort now are easy tasks you do in the course of an hour or two without thinking about it much. Given the giant leap, we should already be seeing the supposed wave of unemployment, right? Companies should be firing most of their programmers because a much smaller number of people can do all the work. Except that’s not what’s happening, they’re building far more code instead. GitHub is seeing orders of magnitude more activity, not vast reductions in the number of people building software. Programming is one of the most obviously transformed sectors of the economy. It should have already been the case that programmers would have been fired en masse. It hasn’t happened and it’s not going to happen. Anyone bringing up horses, or telling me that I don’t understand that the machines will far more capable than any human being, will be laughed at. Of course I understand that they are far more capable at programming than I am already in most respects and will be far more capable still in the future. I was talking about that 35 years ago. I understand your supposed point very well. You just don’t understand economics. If your theory was correct, then we should’ve already been seeing a different outcome.

its literally been 45 minutes...

@KenjiTakano4 @Winterrose let’s go. i have high hopes for the M6 era, especially for the ultra chips.

@BeCachet @Winterrose It's already known. M6 gets eSIM, OLED, and touch. macrumors.com

can we just get eSIM for MacBooks please. this whole “mobile hotspot” feels lame compared to the future I expected by this point

seriously, how do u get rid of these flags? they're in everything single codex session...

a welcoming msg from codex...