Eng @samora_ai (yc w26) , prev @cardlift (yc s24), shiping backend systems, browser extensions & AI Agents at https://teapaw.org/t.co/Qixanwyxvf
Bengaluru, India
Joined April 2022
- Tweets4.9K
- Following1.4K
- Followers8.9K
- Likes13.7K
It's only been 3 days since we launched
Decrawl, and We've got two calls lined up
with founders who are curious to learn more about what we're building
and what Decrawl can do. If you haven't checked out Decrawl yet,
Try : decrawl.sixteenlabs.space/
Every agent I build needs to read the web,
and scraping has always been one of the slowest and most expensive parts of the process.
A single slow fetch can hold up the entire workflow.
So we built decrawl.arisyn.studio
Chrome only kicks in for JavaScript-heavy or blocked pages, and even then, it reuses warm sessions to avoid unnecessary overhead.
One parse gives you everything: Markdown, clean HTML, links, tables, and structured product data. No LLM calls.
I migrated the entire infra of my product from AWS to cloudflare and it reduced my monthly cloud bill by ~85%
We recently launched @beacon_mcp
Which let's your agents control your entire browser and do anything while running multiple agents in your browser in parallel
Try it here : trybeacon.space
Adarsh Jaiswal🥑 retweeted
Launching Beacon MCP 🚀
Your AI agent can now use your browser: your own Chrome, already signed in.
→ Several Chrome profiles at the same time
→ Tasks in parallel, each in its own tab group
→ Your own tab stays yours
→ It asks you when it gets stuck
Works with, any Agent.
Adarsh Jaiswal🥑 retweeted
I built a tool that controls the Chrome you're already logged into.
Works from Cursor, Claude, or a chat in the browser.
You type what you want. It clicks the tab.
This clip: Google Calendar invite, 38 seconds, no speed-up.
Install the extension and try it ,
2,000 free side-panel credits for new signups.
i was working on a product in stealth mode from past 8 months, alongside my job.
and we recently launched it for alpha testing. We got 20 users in one day, and we currently have 17 of them using it actively on daily basis.
Tried it recently and it works insanely well. I’ve started using it for almost every repetitive task.
I’ve used Claude’s extension for a while, but this feels way faster.
If you want your agent to directly control your browser, definitely give this a try.
Adarsh Jaiswal🥑 retweeted
Go 1.27 is shipping with an experimental SIMD API; this is big and super interesting. Some info on this...
Some context: SIMD is a CPU feature that applies one operation (like add, mul, etc.) across an entire vector of values in a single instruction, instead of looping over them one at a time.
For example, to add two arrays of float64, you could loop and do one add per pair. But a SIMD instruction can add 8 pairs at once, packed into a single 512-bit register. So, super optimized.
This optimization is thus important for use cases that crunch large chunks of numeric data like AI inference, cryptography, image and audio processing, etc. Prior to this, the only way to reach that from Go was hand-written assembly, but ... things are changing ... and for good :)
Go 1.27 introduces a new `simd` package, making it super simple to use standard data types while abstracting out SIMD operations. Best part: this is platform-independent.
So, code compiled for a machine with real SIMD hardware runs at near-assembly speed, and the same code compiled for a machine without it still runs correctly, just slower.
Try it out. Hope this helps.
Adarsh Jaiswal🥑 retweeted
Your AI agent needs one table, not your whole database.
I built Client SQL for scoped PostgreSQL access and SQL review before writes.
Free platform access for the first 1,000 users.
Live on Product Hunt.
If you like it, please upvote & share
producthunt.com/products/cli…
Adarsh Jaiswal🥑 retweeted
I built LaunchJury for Claude Day: a jury of launch specialists for the side project you never marketed.
Point it at a repo. It reads the code, researches how similar products landed on HN / Reddit / Product Hunt (real quotes, real URLs), scores your positioning against 8 launch principles, drafts your Show HN, PH launch, X thread, Reddit posts, demo video script and launch video script, then puts them in front of 7 specialist juror personas who predict the top comment your audience will leave and send the drafts back until they pass. Ends with a distribution plan: where to post, when, who to tag.
Built with Claude Fable 5.1 in Claude Code on the Claude Agent SDK. Every stage is one query() call with a JSON-schema output, validated with zod and written to disk. Jurors run in parallel; the foreman is plain TypeScript with unit tests. Fable planned the architecture, wrote the pipeline, and debugged the SDK with me in one session.
First test: I ran it on itself. It scored its own launch 5.1/10 and told me an AI post generator launching to HN readers who hate AI posts is a trap. Fair.
Repo: github.com/NAMANIND/claude-d…
@claudeai @aniketsahu_115 @nosajab @MPSeDC_DST
Mindset shift is real once you start working as a full time software engineer.
The things which used to horrify you now feels necessary.
It all started 3 months back, when me and @Naman_rai_ were discussing ,
what can we build that can become a next generation of automations
and while exploring , one of the yc company drew my attention that was meteor
And guess what, that was insane. We got our backbone of how the agent will understand where to click.
And that's how this got started.
But in the end, cuz of our hectic schedule, we paused it and today @OpenAI launched its own browser and it really does almost everything
But I'm still amazed, didn't this popped up in the mind of brilliant engineers working at @OpenAI . If they thought about it, why didn't they try building it.
Would love to know the reasons for it.
If you're a yc founder/shipper, would love to hear your thoughts as well.