Brady retweeted
I like @unstoppableweb’s multi-agent workflows.
You can chain together prompts & scripts to automate your entire dev lifecycle.
I save so much time with a code review loop that cleans up PRs with both Claude & Codex.
Plus you can use OS models for free.
You can chain together prompts & scripts to automate your entire dev lifecycle.
I save so much time with a code review loop that cleans up PRs with both Claude & Codex.
Plus you can use OS models for free.
3
3
6
2,160
this is why @unstoppableweb has a built-in browser, so you and your agent can give each other some space
0
0
1
287
Brady retweeted
You can use Codex for free with Unstoppable when you download the app: app.unstoppable.ai/code?utm_sourc… #ad
0
3
2
958
Brady retweeted
Interesting approach from @unstoppableweb
Their IDE lets you run Claude and Codex together in code review loops, customize multi-agent workflows, and switch between agents in one place
The cross-agent code review is what caught my attention
Their IDE lets you run Claude and Codex together in code review loops, customize multi-agent workflows, and switch between agents in one place
The cross-agent code review is what caught my attention
14
8
81
6,687
usage caps are part of task scheduling now. before I start 3 agents I’m checking which account resets first and which task can survive getting cut off halfway through.
very senior behavior tbh
very senior behavior tbh
Codex has reached senior engineer level.
Ask it to do something and it regularly replies with "I’m at capacity, try someone else."
0
0
0
563
pointing Codex at the SDK is a pretty great onboarding flow.
now I want to see a plugin with auth, background work and state that has to survive more than one turn. that’s where these usually get weird.
now I want to see a plugin with auth, background work and state that has to survive more than one turn. that’s where these usually get weird.
point your codex at github.com/openai/mcp-ext… to start building a chatgpt plugin!
0
0
0
745
this fixes a very specific annoyance: a Claude session can leave three questions open in three different parts of the transcript, then you get to play archaeology before it can keep going.
an inbox that survives resume and checks things off as they’re resolved is a great idea.
an inbox that survives resume and checks things off as they’re resolved is a great idea.
built a claude code mod called Inbox. it gathers and organizes all the outstanding questions, tasks, and findings during a session and makes them easy to work through.
these items are automatically managed and checked off by Claude when resolved.
it's massively helped reduce n install inbox --marketplace petekp/inbox
source: github.com/petekp/inbox
1
0
3
593
Brady retweeted
I'm really enjoying the Unstoppable IDE 🚀
The @unstoppableweb IDE lets you bring your own Claude and Codex subs, then run review loops with both agents on the same change. Better than trusting one model to review the change.
You can switch between Claude, Codex, and Cursor in inal and browser there too. Planning, bug fixes, and review stay in one place instead of three apps.
The @unstoppableweb IDE lets you bring your own Claude and Codex subs, then run review loops with both agents on the same change. Better than trusting one model to review the change.
You can switch between Claude, Codex, and Cursor in inal and browser there too. Planning, bug fixes, and review stay in one place instead of three apps.
3
4
7
1,871
$400k of codex and five months vs $20k of opus and two weeks is crazy.
but the second run inherited five months of repo state, tests, failed approaches and a much sharper definition of done. I’d love to see what changed before opus took over.
x.com/theo/status/21…
but the second run inherited five months of repo state, tests, failed approaches and a much sharper definition of done. I’d love to see what changed before opus took over.
x.com/theo/status/21…
Announcing tsc-rs (aka ts-rust), my complete rewrite of the Typescript compiler, type checker, and LSP, all in Rust.
This is a project I've had agents working on for 5 months now. I burned ~$400k of Codex tokens and got nowhere. Burned ~$20k of Opus and got there in 2 weeks.
I line of the code.
tsc-rs is an open source drop-in replacement for tsc, and it's available now.
2
0
4
528
A lot of people ask me about being able to connect another server to farm out agent work to while you use your laptop as your daily driver
this is something our team hacks together now and we're working on launching this inside @unstoppableweb soon
we're building out our cloud infra to support dev workloads moving to the cloud, but in the meantime putting your mac mini or old laptop to work is a very cost effective solution.
stay tuned!
this is something our team hacks together now and we're working on launching this inside @unstoppableweb soon
we're building out our cloud infra to support dev workloads moving to the cloud, but in the meantime putting your mac mini or old laptop to work is a very cost effective solution.
stay tuned!
0
0
4
437
Remote control just shipped to @unstoppableweb mobile app
Now you can keep your PRs moving, even from the toilet
Now you can keep your PRs moving, even from the toilet
2
3
8
4,880
This entire thread is people asking cursor to allow bringing their own subscriptions
Even wise news, you won't even be able to use OAI models at API pricing starting mid November
The solution is pick an IDE where you won't get locked into a provider and you can use everything at once
Even wise news, you won't even be able to use OAI models at API pricing starting mid November
The solution is pick an IDE where you won't get locked into a provider and you can use everything at once
@kr0derc@t3dotcodests in cursor are just too good though. you don't need to see threads cluttering up your sidebar if you have a smart coordinator managing them for you. it's entirely changed how i use agents. you'll find that threads and side chats just become busy work for you to manage - when an agent could do it for you
0
0
0
344
One handy @unstoppableweb feature I love -
I hit my Claude 5h limit and it continued on with Opus seamlessly using my Unstoppable AI credits.
Just another reason it pays to stack plans and not get locked into a single ecosystem
I hit my Claude 5h limit and it continued on with Opus seamlessly using my Unstoppable AI credits.
Just another reason it pays to stack plans and not get locked into a single ecosystem
0
2
3
1,041
Brady retweeted
Open four worktrees and paste one of these into each:
"Add dark mode to the settings page."
"Write API tests for the billing endpoints."
"Fix the flaky login test."
"Review yesterday's PR and list what you'd change."
Each agent gets its own worktree, so the four run at once and You come back to four diffs to review.
"Add dark mode to the settings page."
"Write API tests for the billing endpoints."
"Fix the flaky login test."
"Review yesterday's PR and list what you'd change."
Each agent gets its own worktree, so the four run at once and You come back to four diffs to review.
2
3
5
5,089
Brady retweeted
12 months of progress on the jagged frontier:
1) AI got spikier, not rounder.
2) Maths, fluid reasoning, expert work: past human level.
3) There is almost no middle ground. Once a capability crosses human level, it often quickly becomes 100x+ cheaper than a human.
4) Cost keeps lity increases stop. Language and knowledge look saturated but price keeps falling.
5) The gaps are not a budget issue. Human-level memory, self-monitoring and long-run planning are not for sale at any price.
AI is now a superhuman specialist that is sub-human at running itself.
1) AI got spikier, not rounder.
2) Maths, fluid reasoning, expert work: past human level.
3) There is almost no middle ground. Once a capability crosses human level, it often quickly becomes 100x+ cheaper than a human.
4) Cost keeps lity increases stop. Language and knowledge look saturated but price keeps falling.
5) The gaps are not a budget issue. Human-level memory, self-monitoring and long-run planning are not for sale at any price.
AI is now a superhuman specialist that is sub-human at running itself.
143
298
2,728
279,579