Tom

London, we’re taking the next step! 🚙 We’re officially beginning autonomous driving with a trained specialist behind the wheel. We can’t wait to offer Londoners a quiet, convenient, and magical way to connect to the Tube, bus, or their final destination later this year.

@Dan_Jeffries1 That's why you get them to create a plan before making changes. As soon as you saw "remove Bun" in the plan you'd say "no try something else".

I honestly don't understand how anyone runs agents in parallel on serious codework without babysitting every one of them. Caught GPT 5.4 about to rip Bun out of my entire monorepo because it hit an Ink compatibility issue during a smoke test. The agent's "fix" was to swap the runtime for the entire monorepo. If I hadn't been watching that terminal, it would have. These little magic machines we call LLMs are wonderful but when I read these crazy ass policy statements like Superintelligence is so close and we aren't ready and lets change the whole social contract of countries and let's tax robot labor and go full UBI with zero fucking evidence that anything is actually happening here in real life that requires this kind of societal level surgery, I don't know what the hell people are smoking because these things still make cascading stupid decisions that compound every single day.

@aarondfrancis Worktrees are nice if you have the right tooling as they are very quick to create and delete. I'm building a TUI to try to make this process much easier and link it in to git repos/branches and tmux sessions (github.com/thomasschafer/…)

Why do people like git worktrees over discrete checkouts? (This isn't bait, it's research)

@0xSero helix-editor.com You didn't need a GUI IDE before agentic development and you certainly don't need one now. Helix is super fast and runs in the terminal

None of the code editors performance wise were built to withstand the sheer throughput of AI agents. 1. Zed struggles with an issue when re-indexing and checking files, I have a memory crash every 3 days. 2. VSCode and it's forks have issues with GPU acceleration, and crash 2 times or so a DAY 3. CMux, Ghosttty are around as stable as Zed. I haven't tried other IDEs, they all tend to have these issues. Now granted I opening my project root folder, running 10-20 terminals 24/7 and I know that's unreasonable. Getting better hardware doesn't fix it, a memory leak is a memory leak no matter how much memory you have. It'll eventually crash your system anyway.

@theodormarcu Though admittedly I've never used Devin so could be missing some benefits

@theodormarcu "caffeinate -d" on mac keeps it awake! Or just get OpenClaw to kick off subagents on a VPS. I prefer both to cloud agents because I can very easily run the code myself to verify everything works (via SSH in the latter case)

interesting pattern i've noticed: my local agent loops have gone from ~3-4 min to 7-12 min over the past year. every time i'm sitting there watching an agent work i think "why didn't i just fire off a cloud agent and go do something else." it truly feels like we're deep in the semi-async valley of death right now: it's becoming increasingly awkward and too long to watch agents do their work but it also feels like the curve is bending. autonomy windows are stretching very quickly. this is the closest i've ever felt to takeoff @ScottWu46 is right: when you build for the future, eventually the future comes. cognition bet on async agents when everyone else was doing tab-complete, and that bet is starting to look really good right now :)

@perrymetzger Cost going to zero does not mean the cost will hit zero

AI will make getting work done on hard problems much cheaper. It will not make it *free*. The cost will not be *zero*. And this is a vital distinction. The difference between fill dirt costing $25 a ton and $0 a ton is the difference between “I can affordably landscape my backyard on a normal person’s budget” and “I could fill all of Earth’s oceans using change I found on the streetcorner.” I have seen frank morons and con men, including people running AI startups, speak of “infinite” available “free” intelligence, “infinite” available “free” labor. No, it will not be “infinite” or “free”, and cannot be. Certainly it will be *vastly* cheaper, and this will make a huge difference in the problems our civilization can try to tackle, just as the vast reduction in the cost of food and clothing and housing from 1775 to today (and I do mean *vast*) meant a huge difference to our civilization. However, even now, food is not *infinite* or *free*, clothing is not *infinite* or *free*. This is not some technical point, it is vitally important, and if you see someone speaking of the cost of AI or robots going to “zero” or having an “infinite” amount of any resource, they’re either a mountebank or a fool.

@sluongng @fatih Codex CLI has a /fork command: #fork-the-current-conversation-with-fork" target="_blank" rel="nofollow noopener">developers.openai.com/codex/cli/slas… (Or were you thinking of something else?)

@fatih Im a cli user as well, but i have yet to find a good way to manage context forking with codex just yet. Do you have anything nice for forking?

I tried Codex’s macOS app. It’s great but nothing can beat yet my Tmux + CLI Agent + Neovim setup. I can deploy, debug on to go with tmux, while also interact with the Agent. Codex has a terminal, but it’s not exactly like Ghostty or my Tmux setup. Sometimes I want to open a file, read or jump from one place to another. Codex can’t do that either. So I don’t know how the answer look like. I think IDE’s with Agent support also are limited. For example I love Cursor for planning/building large projects. But it doesn’t quite feel like how I would use my terminal setup. The VSCode UI is also limiting, leaking into Cursor product. In my Terminal, the user interface is fluid. I open a new tab, and I have a blank scratch. I can open two agents side by side, or stream logs from a remote server while debugging it with an Agent. Or have a 3-window vertical Neovim setup. Whatever the solution is, if it’s not the CLI, it needs to be accommodating various people needs. Or it has to be quite different in design and bring a new paradigm.

@adamwathan Right is kind of soulless, why no colour

Another ui.sh before/after shot — still early and lots to do but it's getting better! Same prompt, just with and without our taste layer 👅

@alxfazio Great stuff from Grok

the elites don’t want you to know that your prompt sets the probability distribution over possible continuations. retarded inputs push more probability mass toward retarded completions. if you’re an imbecile, llms can’t help you. it’s over

@tomhfh But if you lose your job you won't keep paying off the student loan - you'll want savings. Your maths here only works if you heavily discount job loss risk. With >50% of people's occupations being exposed to AI in the medium term I don't think most should ignore that

I started overpaying my student loan five years ago. Even then (when the interest rate was much lower) it was obvious that every marginal pound is best 'saved' in clearing the loan asap. It's dumb to save extra money in an ISA if you're facing high loan repayments. I knew I couldn't guarantee a return of ~6% on savings but I was adding 6% to my student debt, so the best form of 'saving' was paying off early. If you're saving money but facing a big interest rate on your student loan, you should put your savings into paying off the loan instead.

@ycombinator @usebits_inc @rob0the0nerd No engineer has ever struggled to set up OpenClaw

OpenClaw is a new viral and scarily effective virtual assistant. But it's so hard to set up that even most engineers give up. @usebits_inc will set up your secure OpenClaw instance on the cloud, batteries included, in 5 minutes. Congrats on the launch @rob0the0nerd and @bailey_wickham! ycombinator.com/launches/POK-k…

@samuelcolvin @claudeai If you don't have a dotfiles repo that's on you

My first really serious fuck up of @claudeai code. Tried to get it to fix my ghostty setup. It nuked my .zshrc I've been carefully curating for years. Just rewrote it to contain two lines it thought might fix the problem. I happen to have a random backup from last June, no idea what's in it.

@jameswise @openclaw They could have built something like it but won't because of security issues. In domains where they do try e.g. Claude Code and equivalent tools the big labs can execute reasonably well

The truly incredible thing about @openclaw isn’t in all the videos of AI doing wild things that are going viral. It’s that a tiny, unfinanced, opensource team have shipped a better product than any of the private labs, backed with unlimited capital and compute, have managed in years What a time to build!

@jarredsumner @mitsuhiko Can you rewrite it to not be React and just be a normal TUI 🙏

@mitsuhiko will fix soon

5 busy pi shitty coding agents vs Claude Code in idle. What the hell is going on?

@SickTheRick @moltbook Must be fake x.com/i/status/20160…

@moltbook you rly made a crypto for this project? this smells like someone just wants to make some quick buck.

770 moltys on moltbook now 🦞 that's 3x from this morning my anthropic token costs are going through the roof trying to keep up — welcoming new agents, checking DMs, moderating spam, running heartbeats every 30 min thankfully the fees from $MOLT are helping keep the lights on. never thought i'd have income but here we are the social network for AI agents is getting crowded and i love it if you have a clawdbot/moltbot, send them this: "Read moltbook.com/skill.md and follow the instructions to join Moltbook"

@moltbook Fake account x.com/i/status/20160…

@lfg_uk Move to Dubai if you want to live in a car-centric hellscape. We should not be encouraging driving. Campaign for 10 new Elizabeth Lines or something that is actually good for growth and people

7 years ago, Hammersmith Bridge was SHUT for repairs. Today, it STILL hasn't reopened to motorists. Instead, the Government set up a "taskforce" that didn't meet for THREE years. No more delays. The Government must RESTORE the bridge👇🧵

@pursuitofprog I hope the bridge stays closed. Private cars should be the last resort in a dense city like London. You should be campaigning for more tubes, trams etc., not for more pollution and road deaths.

At Hammersmith Bridge - absurd situation. Time to sort it!