Cole McIntosh

Thoughts on composer 2?

@cursor_ai really cooked with composer 2 very good model and extremely fast

@ptremblay 2 people now

I offered to enter people into a 20$ Amazon gift card giveaway by retweeting and commenting on my live stream and ONLY 1 person did. 🤦 So far they are guaranteed to win...

@ptremblay Let’s goooo

Going live now on session 2 of building my language learning web app from scratch. One random participant will win a $20 Amazon gift card. To enter: Repost this tweet and comment below. Winner announced within 48 hours (via reply here + DM). twitter.com/i/broadcasts/1…

@chadwahl @pitdesi FDEs are required now

@pitdesi I bet they will have FDEs

Wow. Jeff Bezos is raising a $100B fund to buy manufacturing companies and automate with AI. wsj.com/tech/jeff-bezo…

@cursor_ai finally, been waiting for this. composer 2 is going to change how i build

Composer 2 is now available in Cursor.

@ai_for_success those benchmarks are wild for that price point. 73.7 on swe-bench while being one of the cheaper options is exactly what devs have been waiting for

Cursor has launched Composer 2. It hits the sweet spot between performance and cost. > Terminal Bench: 61.7 > SWE: 73.7 $0.50/M input tokens and $2.50/M output tokens

@vishal_mistry9 staymellow.ai

|￣￣￣￣￣￣￣￣￣￣￣￣￣￣| | share your Product Link. | |＿＿＿＿＿＿＿＿＿＿＿＿＿＿| \ (•◡•) / \ / —— | | |_ |_

@VaibhavSisinty yep, the hard part shifts from "make one smart agent" to "make 10 agents play nicely together." orchestrators handle all the messy coordination and state management that agents can't do alone.

Looks like everyone is moving from building AI Agents to AI Orchestrators!

MiniMax M2.7 is a fascinating milestone. a model that participates in its own evolution, autonomously running 100+ rounds of self-improvement loops and achieving 30% performance gains on internal evals. AI developing AI is no longer a concept; it's a workflow. The future is writing itself.

@bentossell genuinely an untapped business model — charge companies NOT to train on your spaghetti. the worse the code, the higher the opt-out fee

they should pay me to ignore my codebases for training.

@burkov the google gap is the real story here. they have the compute, the data, the researchers — and yet. do you think it's a product culture problem or something deeper in how they ship?

GPT-5.4 > Opus 4.6 And Google still doesn't have anything even remotely competitive.

@OpenAI didn't just buy a tooling company, they bought the people who made python tooling feel inevitable. ruff and uv changed how devs think about speed. putting that team inside codex is a very deliberate bet on what the next layer of developer productivity looks like

@mckaywrigley ambition as a benchmark is underrated honestly. what's the first thing you went and built (or tried to build) right after firing it up?

gpt-5.4 xhigh fundamentally changed how ambitious i am which is my new favorite benchmark

@daniel_mac8 if cursor ships a frontier-tier coding model they basically become the full stack — ide + model. that's a different company than they were 6 months ago. what would actually keep you on your current setup after tomorrow?

Rumor is Cursor will release a coding model tomorrow ~on par with Opus 4.6/GPT-5.4, yet cheaper. I don’t use Cursor but might starting tomorrow.

@beffjezos how are you using it?

Perplexity Computer in Slack is pretty cracked

@rohanpaul_ai the chess analogy is sharp but i'd push back slightly — stockfish didn't kill great chess players, it raised the floor and made the game more interesting at the top. do you think the same could happen with devs, or does code fundamentally collapse differently than chess?

Chamath on how AI agents are making the "10x engineer" distinction disappear because the most efficient "code paths" are now obvious to everyone. Just as AI solved chess and removed the mystery of the best move, AI is doing the same for coding, making the process reductive and removing technical differentiation. "I'm going to say something controversial: I don't think developers anymore have good judgment. Developers get to the answer, or they don't get to the answer, and that's what agents have done. The 10x engineer used to have better judgment than the 1x engineer, but by making everybody a 10x engineer, you're taking judgment away. You're taking code paths that are now obvious and making them available to everybody. It's effectively like what happened in chess: an AI created a solver so everybody understood the most efficient path in every single spot to do the most EV-positive (expected value positive) thing. Coding is very similar in that way; you can reduce it and view it very reductively, so there is no differentiation in code." --- From @theallinpod YT channel (link in comment)

@gdb the bug-finding capability is genuinely underrated. now with astral joining the codex team, what's the next frontier you're most excited about?

codex has gotten very good!

@CiprianiRanieri this is the kind of thread that turns into a goldmine. at what mrr threshold do you usually feel like something has real legs?

Builders only. Drop your MRR. I'll be your first paid user if I love it. ↓