Mohsen Azimi (@mohsen____) header
Mohsen Azimi (@mohsen____) profile picture

founder @prooflinedev, previously @airbnbEng, @elevenlabsio @lyftEng, @googlecloud cache ruins everything around me

Joined April 2009
Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)Photo by Mohsen Azimi (@mohsen____)
How to grow in ChatGPT - build a product is useful - open the MCP for ChatGPT to see, suggest, and use - make agentic onboarding easy for agents and effortless for humans - build a good product that is useful - just do the right thing
53
This is new, OpenAI lets you buy resets for Codex
1
121
CI/CD in TypeScript... what can go wrong?
Replying to @harjotsgill
3. CI/CD: a DAG compiler, scheduler, isolated workers, fenced artifact publishing and recovery. 36K pipeline runs and 1.7M+ jobs a week. Typical PR runs: 4–5 min. 1,312 PRs · 151K TypeScript LOC.
153
I never understood why anyone would pay for dictation. I was using Apple dictation before starting to use Codex (ChatGPT now?) app. Just double tap right option key and go!
Wisprflow death clock 🕰️ 🪦 How long until $2 billion goes to zero?
1
117
I like API docs that let me copy the first example, run it in my terminal, and see my own data come back. So we added a little shortcut to the Proofline docs. Pick your organization and generate a read-only test token, and the examples fill in with your details. Then copy the curl command to pull up your latest reviews. The Python and TypeScript examples work this way too. Here’s what that looks like in 42 seconds
1
1
85
i broke GitHub. It can't show my contribution heatmap anymore 😅 github.com/mohsen1
1
1
77
AI can produce text quickly, but producing text isn’t the same as developing an idea. Even when authors control the prose AI generates, a finished passage can bypass the decisions that clarify their thinking. Writing makes us decide what we mean and how another person will come to understand it. Great writers also map the ideas they want to convey, considering what readers know and introducing ideas in an order they can follow. AI can help with that work without replacing the creative writing process. It can help authors draw on principles of thinking and writing developed over centuries, including the practice of considering a reader’s understanding as ideas unfold. That’s why I built Wercend. Rather than simply suggesting the next word or paragraph, it helps authors see the structure of their ideas and stories, notice missing connections, and recognize where readers may need more context. Wercend is a modification of the Old English word Wyrcend, meaning ‘worker.’ It won’t tell you what to write. Instead, it helps you do the work of shaping your prose: examining its flow and how one idea or event leads to another, so you can decide how best to guide readers through your story. wercend.com
51
good prompt: do an audit to find all state managements that could've been avoided entirely in my app
2
162
Been using the touchscreen MacBook Pro for a full day and I have touched the screen exactly one time so far. I was testing to see if it's working actually!
1
99
For @prooflinedev we use GitHub projects. Because at the end of the day it’s a glorified database for tickets. UX doesn’t matter
our team uses linear, but everyone just connects their mcp and chats with it. i used to love linear because it was an amazing product experience and so pleasant to use but now it's weird to pay for 6 seats on linear when we just use it as a glorified database. at this point, it's hard to justify paying for it
1
1
116
Sonos app takes 9 seconds and 4 taps to change which rooms are playing. and thats when it works so i made my own. runs on my router. 2 taps, under 2 seconds open source. works on gl.inet, dd-wrt should too but havent tried a real one github.com/mohsen1/rooms-for…
2
1
6
417
Two of my favorite technologies: @OpenApiSpec && @typescript
SDK inference from OpenAPI files has been achieved internally. Finally, a full no codegen SDK generator.
3
158
Mohsen Azimi retweeted
Focusing on sandboxing is fundamentally the wrong idea. A perfect sandbox is useless because whatever you put in the sandbox doesn't have access to any data. So instead of opaque, hard boxes people use transparent, soft ones, and the problem then becomes the composition of transparent boxes and the management of permeability. Except that there is no legitimate theory behind this operation and the mechanisms are always abstruse and ad-hoc and misunderstood by users. The only real solution is to think about data access and provenance, and use pure computational processes, not boxes.
6
8
1
53
6,813
> when asked to fix one bug, it can identify the broader class of problems behind it and ask whether it should address those too This model will also suggest to shut down the whole thing and go find better things to do with your life theinformation.com/newslette…
72
Mohsen Azimi retweeted
Deterministic simulation testing should be part of every conversation with your coding agent. Otherwise, you’re doing it wrong™️
8
3
3
57
6,583
why it says byte-for-byte and not bit-for-bit?
70
Mohsen Azimi retweeted
in the Think post, I proposed this idea of an execution Ladder. Where the LLM (or even the user) _chooses_ to escalate the execution environment based on what it's trying to do at the moment. I'll make it so this isn't cloudflare specific, you can bring your own browser/sandbox/memory service. but it implies an impending feature/capability: those execution environments might have their own harnesses, and Think will have to communicate/work with them. indeed (I mentioned this in my code mode talk) _users_ will be communicating to Think (and other apps) with _their_ own harnesses, running on their phones and whatever. I suspect some form of standard will arise for cross-harness communication/collaboration. blog.cloudflare.com/project-…
9
14
3
137
15,346
⁠Will Instinct raise funding at a valuation above $10 billion before December 31, 2027? @Polymarket I want to see this 🍿
60
VCs: "we're good at finding patterns in this market" The patterns: "are other VCs investing?"
2
96