
Founder of @Streak, previously @Google, @ycombinator
San Francisco, CA
Joined April 2008
- Tweets2.9K
- Following424
- Followers1.6K
- Likes1.3K
Congrats to @JAXA_En on MMX! By returning samples from Phobos, Mars’ larger moon, the mission will add to decades of Mars exploration and help shape future discoveries and potential crewed missions to the Red Planet.
[Press Release]
Launch result of the Martian Moons eXploration (MMX) spacecraft aboard H3 Launch Vehicle flight No.10
global.jaxa.jp/press/2026/10…
Aleem Mawani retweeted
In 1983, a BBC interviewer asked Richard Feynman why two magnets push each other apart — and he refused to answer it.
What he did instead is the best seven minutes on thinking ever filmed.
Bookmark & watch today, no matter what.
is anyone actually building "code mode" style MCP servers in production? Does it actually work?
Cloudflare (blog.cloudflare.com/code-mod…) swears by it but wondering how it works in practice.
how do you trust your agents' work more? copy this post/link and give it to your @bot or agent!
1. codebase improvements: this is just good software engineering - better architecture, stronger types and invariants, clear boundaries. agents love to copy what exists and take the shortest path, so make the easy thing the right thing. categorically eliminate problems through code and you won't need to think about it again.
2. static analysis: lint rules, tests, CI, verification (eg TLA+). very token efficient ways to catch issues ahead of time.
3. prose: comments, rules, AGENTS.md, skills - anything that can't be enforced and requires agent judgment. use it minimally, but do use /create-verification-skill.
4. human review: the hardest to scale. at the rate agents land code, it's pretty much impossible to thoroughly review every PR, so spend that effort on the highest leverage point, designing an environment where agents do the right thing by default.
this forms an inverse pyramid: you want most of your effort at the top, least at the bottom. codebase improvements are deterministic, making it more token efficient, and lets you sleep more soundly at night.
install github.com/cursor/plugins/tr… and create a new Project:
1. look through my slack/chat messages and code review comments where i've given feedback on PRs, code, architecture
2. use /recall on my past transcripts to learn how i steer agents
3. /poteto-mode use /correct and /architect on past commits to find the most common pitfalls agents fall into, then propose a plan to rearchitect the code i own so it's more agent friendly. answer open questions with prototyping rather than defer to me, and back the plan with real data.
4. before you begin, use /sharpen + /in-your-own-words and confirm this prompt with me
Aleem Mawani retweeted
The thing I’m most grateful to Elon for is bringing back an age of wonder and excitement to humanity. That was jaw-dropping awesome! Screams and mouths wide open in the audience. Nuts.
Aleem Mawani retweeted
Replying to @Scott_Wiener
Scott, the profits already are taxed. Twice.
(1) Corporations are taxed on net profit. If they make more profit from these AI efficiency gains, government gets more income
(2)those profits are again taxed when distributed as dividends.
Now government should make sure Kroger has competitors. Which they do. And then ultimately these AI efficiencies across the industry will both benefit consumers with lower costs AND drive more corporate taxes.
Efficiencies are good. Remember when we had to pay bills by handing a dude a piece of paper and ask him to drive it across town in a mail truck? Isn’t it good we have internet payments now?
Aleem Mawani retweeted
Selling only to indie makers is the reason you have a high cancellation rate in your software business. This was already the case before AI but it's accelerated now because they love building & hosting their own stuff to save small amounts of money. Sell to businesses. Trust me.
Aleem Mawani retweeted
I believe the "we ship 30,000 changes/month" stories. But I suspect the type of product you're shipping into is important.
If your product is like a big garden with 500 trees, each can be trimmed on its own. Integrations, tools, connectors, edge-cases, security fixes. Each change makes one tree better, so tons of changes can run in parallel.
Other products are more like a bonsai. Every cut changes the whole thing. If I want to cut here and you want to cut there, we have to actually agree first.
I wonder if a lot of the "code review" debate might be apples and oranges depending on whether you're a platform or a bonsai team.
A formalism of this is Kauffman's NK landscape. N is how many decisions you can make. K is how entangled they are. A wide-surface product has large N but small K. So it can absorb enormous parallel change. A tightly integrated UX has high K, so parallel changes collide in what they're trying to achieve, even when the code merges cleanly.
Aleem Mawani retweeted
My daughter's middle school has the most direct fundraising strategy and I'm here for it.
We can skip the whole charade.
Aleem Mawani retweeted
Replying to @JamesWard
I love the simplicity of Go. For a human world, it's a beautiful design. The feedback loops are fast (it blurred the line of 'compiled' vs 'interpreted' for many projects). The toolchain is so simple. The code is easy to follow.
AI is def making software @Streak better.
I now do AI powered UI reviews where the AI is fixing every little tiny detail I find. I would never torture an engineer the way I'm torturing AI to be absolutely pixel perfect with exacting animations.
If Apple can’t make Siri better they should at least have the decency to let apps like @WisprFlow integrate more deeply. Having it be a 3rd party keyboard is such stupid UX
✍️ Writing is essential to how @Mercury communicates as a 1300 person remote-first company.
I wrote on Writing At Mercury (inspired by @vxanand's post).
Principles of Writing at Mercury
1. Writing is thinking: Spending time on writing is not wasting time, it helps you process ideas, analyze problems, and explore solutions.
2. Less writing is better: Good writing is concise, which makes it easier to understand.
3. Reading should be easy: If >2 people will read your writing you should spend extra time making your writing good + digestible + clear etc. It is not fair for people reading to do work that the writer could have done for them.
4. Writing is expression: Your writing will be taken as an expression of your identity, so give it the care it deserves and be authentic.
5. Writing is important: Writing is the main way we communicate internally at Mercury, as a remote-first firm. We frame important discussions, and document thought processes and ultimate decisions, with our writing.
On Using AI for Writing
It is very easy to use AI to write a lot of content. If you do so, then by default you have:
1. Outsourced your thinking
2. Created a verbose/long doc
3. Made it hard to read
4. Have a doc that is not expressive and very bland
5. Watered down our discussions and decision making
In other words, heavy AI use for writing is diametrically opposed to the principles of writing laid out above.
Some effective ways that AI could be used instead:
1. Critiquing your writing
2. Condensing your writing
3. Shaping initial drafts
4. Summarizing complex information
On Using AI to Respond on Slack/Comments
Slack threads and comments are a particularly bad use of AI writing.
1. Slack responses and doc comments should always be short. Large blocks of AI text don’t belong.
2. AI used here is often thoughtless and the other party could have just used the AI tool themself. It inherently shows a disrespect for the reader’s time.
3. Given our remote culture, one of the main ways we connect with each other is via Slack. Using AI often would obscure your authentic perspective.
4. It would be better to share the prompt you gave to the AI directly to the other person and they can just prompt their own AI tool.
How to do Great Writing at Mercury
We aren’t going to police whether someone is using AI or not to help them write. We will just have a high standard for your writing output.
Here are some high-level rules that should help:
1. Always read and stand by 100% of your written output.
2. When it is important, spend time on your writing. You might spend time on the narrative arc, the tone, the structure, whether it clearly articulates an issue, explores solutions, and drives toward an outcome.
3. Think hard about how you can use fewer words and be more concise.
4. Be authentic. You don’t need to dress up your writing in corporate jargon to make it sound professional. Simple, plain writing is often more effective.
5. Never copy-paste AI responses to someone in Slack or to comment on a doc.
6. Sometimes it's better to write nothing at all. Don’t make a comment in a doc/slack because you want to be part of the convo, say it if it materially moves things forward.
Been playing with Jev and already shipped a prompt classifier for @Streak to production.
Being able to do intelligence *as you type* opens up some awesome UX.
Our median latency from 1.5s to 200ms and a 74% increase in our core metric for the feature!
Aleem Mawani retweeted
TI-83 BASIC while bored in class, iykyk
Fun story: I made a prank calculator that gives subtly wrong answers, to install on my friends’ calculators… and then accidentally used it myself for a chemistry test
Aleem Mawani retweeted
We got Grok Bot in our Teslas before the iPhone had a functioning AI assistant.
Replying to @danieldbunn
I'm sure someone in government is working on solving that travesty of justice :).






















