Founder @mesa_dot_dev. Prev: YC, Census, Microsoft, Northwestern
San Francisco
- Tweets 3,213
- Following 1,097
- Followers 1,637
- Likes 39,513
Pinned Tweet
I'm so excited to officially share what we've been cooking up at Mesa: the most powerful filesystem ever built for AI agents.
The dirty secret of every "production" AI agent today: the filesystem is held together with duct tape.
Teams are stitching together S3, GitHub, sandbox-local disks, and homegrown diff logic to give their agents something resembling persistent, versioned storage. None of it works.
S3 isn't designed for parallel agents - concurrent agent writes silently overwrite each other.
GitHub has the semantics but rate-limits you into the ground at agent scale and doesn't give you filesystem ergonomics.
Sandbox disks vanish the moment the container dies.
And your agents don't want to git clone and git push anyway. 𝗧𝗵𝗲𝘆 𝘄𝗮𝗻𝘁 𝘁𝗼 𝗿𝗲𝗮𝗱 𝗮𝗻𝗱 𝘄𝗿𝗶𝘁𝗲 𝗳𝗶𝗹𝗲𝘀. 𝗟𝗶𝗸𝗲 𝗲𝘃𝗲𝗿𝘆 𝗽𝗿𝗼𝗴𝗿𝗮𝗺 𝗲𝘃𝗲𝗿 𝘄𝗿𝗶𝘁𝘁𝗲𝗻.
So we built the missing layer.
Mesa is a durable, POSIX-compatible filesystem with version control built in. Branches, diffs, history, rollback, access control — every primitive a codebase has, for any file type, at agent scale. You mount it. Your agent uses it like a normal filesystem. We handle the rest.
Private beta is live. Link in comments.
36
26
333
48,429
The real reason CS people are winning is because internet and AI were tech tree breakthroughs
But there's something to be said for coding letting you train general systems design skills at a cycle time that's impossible in something like civil engineering
The reason CS people seem to eat every industry is because it's purely the study of time complexity in problem solving
5
236
If your restaurant makes lasagna with ricotta instead of bechamel you should be shut down immediately
1
201
Greatest returns from exile:
5. Tom Brady at Tampa Bay
4. Steve Jobs at NeXT
3. Napoleon on Elba
2. Moses in Egypt
1. @ElizabethHolmes at Federal Prison Camp Bryan
6
568
I expect many companies to start sending RL envs to the big labs for free
Imagine you're building a CRM. If Claude is great at using Salesforce, but sucks at using your platform, your product is effectively much worse
Basically souped up SEO
Just spoke to one of the big data labeling businesses.
Few interesting insights:
- They predict the majority of their revenue will come from Fortune 1000 enterprises, not labs in a few years
- They believe every company will want to own their intelligence, but owning intelligence does not necessarily mean using open source models
- A company’s evals will become their main proprietary IP given the improvement in agent performance after properly setting up & running internal eval environments
- Most enterprises haven’t graduated from coding agents and it’s largely due to not having the proper eval infrastructure to make non-Eng agents performant
7
412
llm writing is so distinctive that you can even tell when someone clauded the script for a youtube video
12
228
At what point in the ai wave do we have “firing bonuses”?
“Hey you literally automated yourself out of a job, here’s a million dollars”
5
11
1,388
agents are finally good enough for normies to use
they used to fail at basic tasks often enough that you needed to let people pick the model, add skills, display tool calls, etc. This meant pretty complex ui. not noob-friendly.
but as we're seeing with grokbot, instinct, meta muse, you can now hide most of these knobs from users and still get good results
2
7
329
Everything about building a company -- setting a vision, holding a product roadmap, maintaining a code base -- is a constant fight against entropy
1
7
212
big tech engineers are good at coming up with solutions
but where the big tech types can fall short in startup world is focus --with limited resources, it's essential to quickly identify which problems are worth solving, and which can just be ignored
1
1
16
695
a few of the chinese ai labs would've paid a friend 7 figs+/yr for his software + training data
he won't do it on principle
meanwhile most tech bros will sell their souls over a coffee chat
A buddy of mine built a super high fidelity Excel clone (bug-for-bug compatible) - runs on Linux/Windows/Mac
there's an angle to sell the software/data as part of an RL env, but he doesn't have good lab connections
anyone buying this kind of thing or brokering these deals? he would give a very large commission
DMs open
1
7
1,980
Don't let AI slop your docs
Even internal docs like your README
Even if AI is the primary consumer of the docs
Any document that confuses a human will also confuse an AI
2
7
649