Worlds best coding agent is a north-star internally at Reason Machines. We have had several moments internally where this has been close. Regardless, building a valuable product and interface to use this in is just as important, always staying at the top For the fastest moving teams. @reasonmachines
3
240
following the hype, deleted ~860K lines of unit tests in our monorepo @reasonmachines the product is very self sufficient and agents do indeed write a lot of unnecessary tests. let see how this goes
5
260
I got on twitter for the first time after attending the xAI hackathon last year. Event that resulted in 10~ hires, 1 remark, 2 dead startups And a talent pool that I meet at every event till this day. Since been back in sf 3 times for research and yc and more. Elon energy flow
6
53
1,472
The frontier will only be getting better and we think harnesses will matter a lot in optimizations for cost and quality. We are launching a lot of things tomorrow @reasonmachines
4
30
758
My software factory does the same thing. Without a terminal mess and local loss
my software factory uses 5-10B tokens a day 24x @DevinAI SWE-2 Max 6x Codex Astra Ultra 2x Claude Code Fable Max running in @zeddotdev
2
13
974
Majority of my day i don't code anymore, yet I still code more than ever Onboarding Engineers all over SF Wherever you are I will come to you. There is no real Reason (pun intended) to be working at a pace slower than this
2
14
638
pun - this is funny because the company is called Reason Machines and the autonomous software engineer is called Reason Agent btw
3
152
regardless, this is the best foreseeable outcome
I don’t think Jev will survive. @nvidia and @meta will release open-source Jev competitors within the next 30 days. @OpenAI will add a Jev-like endpoint to their model list in the next 60 days. And the rest of us will fine-tune custom classifiers that crush Jev and are 10x faster. In two years we’ll look back at this flash-in-the-pan and say “remember Jev?”
2
6
582
Amazing sites built with Reason. Right in chrome extension.
2
1
12
606
The worlds best coding agent should have these things: A file viewer, Remote coding, Cloud coding BYOK and Sub Computer use and browser use A full codebase viewer (PRs etc) All 3rd party integrations Fast, native Excellent multithread handling (A2A) And a good harness Who is it
4
9
551
We are a two people team that are building the worlds best multiplayer coding agent. If you could tell me that we could outship frontier harnesses i would say disagree. Would need ~1 OOM more tokens for that. Things need to either get cheaper or we increase budget
3
1
18
765
If personal intelligence agents will owned by everyone (muse/grokbot/instinct) then what will other agents do? World class domain agents will be supplied as mcps to them. Thats my prediction
11
325
Reason Machines aims to maintain the fastest moving infrastructure. Running secure agents at scale is hard, and @blaxelAI has been tremendous to work with. Congrats to Paul, Nicolas, and the team on getting acquired by @baseten. blaxel.ai/blog/reason-machin…
1
12
422
Adi Singh retweeted
.@reasonmachines built an autonomous software engineer that opens and merges PRs without the developer ever touching the app. That requires a runtime that matches the model: sandboxes that resume in 1 second, hold state between tool calls, and auto-suspend during reasoning gaps. How they built it on Blaxel: blaxel.ai/blog/reason-machin…
2
12
1,065
Today is YC Demo Day for YC S26. Watching everyone take the stage right outside our place brings back how fast things move. We were pre-dispatched earlier, and since then we have stayed purely focused on building Reason. Instead of coding agents breaking in local terminals, we wanted entire teams running agents in cloud infra with auto-generated PRs and visual screen recordings to verify every run. Excited for every founder pitching today. Building the future at reasonmachines.ai
3
41
1,554
computer use and browser use is terrific now (even on open source). the bottleneck to making machines entirely autonomous is native OS permission gating (primarily Apple) and custom popup machinery. it feels like the legacy bounded gates of our time.
5
10
494
A good evaluation suite is the difference between a decent agent and a stellar one. I think the labs have got this. And is a majority of the reason a product "feels" better. It's backtested
2
7
442