Andrei Otel

Solo building a project board for people who code with AI agents. The idea: your agent is a teammate on the board, not a black box in a chat window. More soon.

@FCB_Cartel @sama getting the opus 4.7 experience I see, its coming full circle, can't wait for 4.8 to be great for 2-3 weeks

Codex in the last 3 days has been a nightmare. Hallucination after hallucination, writing so much unwanted code, creating things that are not at all intended, taking too much time and burning tokens unnecessarily. Especially /goal. @sama something is wrong. It burned 50% of my Pro plan tokens for the week in 1 night creating a PR of size 104k lines and when asked about the PR : it said it’s useless and cleared the whole PR and closed it.

@camiinthisthang Design and product vision 100%. As a tech solo founder I find myself going to design and product friends for feedback more than I should…

I’m more and more convinced every entrepreneur needs their cofounder to be a cracked designer

@t_blom I want to hand-hold initial customers, how else am I going to know what makes sense for them and what doesn't

Biggest red flag I hear from early stage founders? “We’re working on self-serve onboarding” Your product currently requires you to talk to every new customer? That’s a good thing.

@grimcodes It is always worth it even if it feels like wasted time. The next day I feel more confident to keep on vibin on a solid codebase

just spent like 10 hours on a pr just to make my code a little prettier

@buildwithsid im just afraid that having Max turned on will just degrade the output since more thinking on an unpredictable model can put in more effort in building the wrong thing.

nowadays i've been using opus 4.7 max for the smallest tasks possible - fix this div - change this color - make it bigger surely i'm not the only mf doing this

@GrizzShutsDown looks really clean, when picture in picture mode is enabled it has additional controls which are not enabled on the main player, not sure if its intentional for demo purposes

When i was creating my website i couldn't find a simple well designed video player so i created mine, I decided to open source it and spent the weekend working on it. I hope you enjoy it like i do, simpleplayer.grizz.fyi

@sugandhanisa It asks for code review week, every 2 weeks you go back and cleanup the crap it generated

Trying to build too much too fast with AI actually slows you down.

@bcherny My biggest problem with CC is the inconsistency in it's decisions. I am askig for research, it starts doing it then implementing based on it, other times I ask it to plan and it requires 2 approvals before starting doing changes. Dang skip perm btw

People often ask what my biggest tip is for getting the most out of Claude Code. These days my #1 tip is: use auto mode Auto mode means no more permission prompts. It is the key building block for multi-clauding: start a session, then while it runs, work on another session in parallel.

@testingcatalog just release old opus which followed common sense and didn't need hand holding

ANTHROPIC 🔥: Mythos 1, "claude-mythos-1-preview", is being prepared for a release on Claude Code and Claude Security. The model became visible for a short amount of time on Claude; besides that, new strings mentioning Mythos have been added. > Access to the Claude Mythos model in Claude Code and Claude Security. It still doesn't mean the general public will have access to this exact model, according to Anthropic's earlier communication. More below 👇

@KellehEyad @cursor_ai That sounds fun, would be really interesting to see how far you can go, given the tkn/s speeds. Sign me up!

Shall we host an online hackathon together @cursor_ai , building only with Composer 2.5?

@kobyjconrad At this point it's better to have multiple 20$ subscriptions to different providers rather than a max plan and be vendor locked

Low key - last thing you want rn is vendor lock in The new meta is rotating every 3 months Cursor -> Claude Code -> CodeX -> Cursor This is the way 🥹

@ClaudeDevs had to double check the date of this tweet

Two updates to auto mode: · Now available on the Pro plan · Sonnet 4.6 is now supported, alongside Opus 4.7 Shift+tab, and let Claude run.

@blader The amount of times I started using caps lock... It's driving me crazy

honestly claude code / opus 4.7 feels barely usable compared to codex / 5.5 at this point for any kind of long running / complex engineering work

@sergeynazarovx @cursor_ai Comfort sentence starter

Why does Composer 2.5 add Chinese hieroglyphs in the beginning of the answers? @cursor_ai

@bcherny Please share the babysit agent

In the next version of Claude Code: run /usage to see a breakdown of which Skills, Agents, MCPs, and Plugins are using your tokens CLI today, coming to Desktop next

Learning so much from releasing to a close group of friends. Thought they'd be the easy crowd, turns out everyone has wildly different points of view. Who would have thought... Genuinely the best thing I've done.

@sama Right after Karpathy solved a+b leetcode for the Anthropic interview? Coincidence?

a general-purpose model solved a major open problem in mathematics. we'll be saying this a lot over the coming years, but this is a kinda big milestone. i'm very excited for AI to greatly extend our understanding of the world, but still, i have complicated feelings today.

@peterpme @grok give me a prompt to scan for these concerns

"React keeps re-rendering. The app is slow" Yeah because you recreate half the app on every render. You underestimate how important stable references are when it comes to React performance.

@rtroar meta changes every week

every day we the SWE role becomes more like playing StarCraft

@Miles_Brundage just use agent teams in claude, it spins multiple agents each with its own worktree, you'll be out of quota in no time

Protip - if you wanna really burn through your Codex and Claude quotas with reckless abandon, specify a minimum number of sub-agents to use