sarath menon Profile
sarath menon

@sarath_suresh_m

Followers
173
Following
1K
Media
67
Statuses
301

Building Mix, the production ready agents SDK cofounder at https://t.co/KOcTKF5RZI Master in Robotics at ETH Zurich, Switzerland

Zurich, Switzerland
Joined May 2023
Don't wanna be here? Send us removal request.
@sarath_suresh_m
sarath menon
2 months
You can't build multimodal apps with Claude Agents SDK. No video/audio tools, local only, single model, no devtools. Mix SDK fixes all 4 issues + deploys to Supabase in one command. Watch: portfolio analyzer + video search agent built with Mix and Supabase → MIT licensed:
3
10
34
@sarath_suresh_m
sarath menon
8 days
This is exactly what happens when you 10x code velocity without 10x-ing testing. AI coding tools gave us superhuman speed. Testing Agents give us superhuman quality control. 85% bug detection vs 30% is a game changer
@nicolasmunty
Nicolas Muntwyler
8 days
Today we are launching Testing Agents AI coding tools like Cursor help you ship a lot more code... and a lot more bugs. Testing either gets skipped or simply can't keep up. Let's fix that. In your Pull Request, just tag our Testing Agent and tell it what to test. It will read
0
0
3
@pederzh
pederzh
14 days
in 🇮🇹 we grow up with very high standards
@Court_Reinland
Court Reinland
14 days
@IterIntellectus My brother in Christ, who are you calling mid?!?
0
1
4
@sarath_suresh_m
sarath menon
24 days
20 hours since launch our salary checking app 320+ people already used it to check if they're underpaid. wild how many people don't know their market rate. check yours: https://t.co/XY5V4uKmLc
@sarath_suresh_m
sarath menon
25 days
nobody talks about salary. What if you're underpaid ?. We built something to fix that. enter: role, experience, city, company tier. AI tells you: •⁠ ⁠market range for your profile •⁠ ⁠where you fall •⁠ ⁠if you're underpaid (and by how much) tested it for
0
0
5
@sarath_suresh_m
sarath menon
25 days
nobody talks about salary. What if you're underpaid ?. We built something to fix that. enter: role, experience, city, company tier. AI tells you: •⁠ ⁠market range for your profile •⁠ ⁠where you fall •⁠ ⁠if you're underpaid (and by how much) tested it for
0
0
6
@sarath_suresh_m
sarath menon
29 days
Built a resume roaster in 30 minutes with Nova. Analyzes your resume against the job description, and roasts every weak point. The "backend"? Three paragraphs of natural language. Try it here: https://t.co/yAVLIatWf1
2
0
5
@sarath_suresh_m
sarath menon
1 month
Here's the entire "backend" for the YT → Thread converter built in Nova All app logic is in a 3 paragraphs. No code. This would be 500+ lines in Express/FastAPI
@sarath_suresh_m
sarath menon
1 month
Built a YT → Twitter thread tool with Nova The "backend" is literally 3 paragraphs of instructions. No API routes. No controllers. No debugging. Join waitlist: https://t.co/hK7Kl42uKy
0
0
5
@sarath_suresh_m
sarath menon
1 month
Built a YT → Twitter thread tool with Nova The "backend" is literally 3 paragraphs of instructions. No API routes. No controllers. No debugging. Join waitlist: https://t.co/hK7Kl42uKy
1
0
6
@sarath_suresh_m
sarath menon
1 month
After posting the this demo, my DMs exploded. Everyone asking the same thing: "How did you build it?" Here you go, live build in Nova
@sarath_suresh_m
sarath menon
1 month
Video fact-checker build on Nova: 100% agent-powered architecture. No backend logic. No complexity explosion. Just: Video URL → Agent prompt → Verified claims with sources. Join waitlist: https://t.co/XjboCZjPnZ
0
0
4
@sarath_suresh_m
sarath menon
1 month
Tired of @claude_code making backend changes breaking your frontend? I use Orval to auto-generate types from OpenAPI for both sides. One make command. Done. Then I told Claude in CLAUDE md to always update the spec and regenerate types after API changes.
0
0
1
@sarath_suresh_m
sarath menon
1 month
The craziest part about that last Nova demo? Traditional approach: Change the output format = rewrite the handler function, update the response serializer, rebuild, test, deploy. Nova approach: Change prompt to "Show this as markdown instead."
@sarath_suresh_m
sarath menon
1 month
Nova demo 2: I just asked it to show output as a markdown document instead of a table. Entire "backend change" was editing a prompt. That's it. This is what happens when your backend is natural language Joit waitlist: https://t.co/XjboCZjPnZ
1
0
5
@sarath_suresh_m
sarath menon
1 month
One exception to this rule: If you're building with Claude's Agents SDK, use the TypeScript version. The only other option is Python. Worth the TypeScript tradeoff.
@sarath_suresh_m
sarath menon
1 month
Avoid TypeScript backends at all costs. If you absolutely must use TS on the server, use Bun. Node + TypeScript = pain. Bun makes it bearable.
0
0
2
@sarath_suresh_m
sarath menon
1 month
Nova demo 2: I just asked it to show output as a markdown document instead of a table. Entire "backend change" was editing a prompt. That's it. This is what happens when your backend is natural language Joit waitlist: https://t.co/XjboCZjPnZ
1
0
3
@sarath_suresh_m
sarath menon
1 month
Avoid TypeScript backends at all costs. If you absolutely must use TS on the server, use Bun. Node + TypeScript = pain. Bun makes it bearable.
1
0
2
@sarath_suresh_m
sarath menon
1 month
Three things Nova won't be: ❌ Another landing page builder ❌ A Supabase wrapper with chat ❌ A "better Lovable" We're building the platform for agent-powered applications. joim waitlist:
Tweet card summary image
nova.recreate.run
Nova
0
0
4
@sarath_suresh_m
sarath menon
1 month
We just shipped a feature that rejects customers Nova now politely declines to build landing pages and simple CRUD apps Just Use Lovable or Base44. They're great at it.
0
0
1
@sarath_suresh_m
sarath menon
1 month
Who's using tanstack DB in production ?
1
0
9
@sarath_suresh_m
sarath menon
1 month
Your Lovable app hits 50k lines and becomes unmaintainable. Your Nova app stays at 200 lines because agents handle the complexity. This is the difference between generating code and orchestrating intelligence.
@sarath_suresh_m
sarath menon
1 month
Video fact-checker build on Nova: 100% agent-powered architecture. No backend logic. No complexity explosion. Just: Video URL → Agent prompt → Verified claims with sources. Join waitlist: https://t.co/XjboCZjPnZ
2
0
11
@sarath_suresh_m
sarath menon
1 month
I was wrong about @vercel AI Elements. Thought it was tightly coupled to the AI SDK. Turns out it works beautifully with anything. Props to @haydenbleasel for building it the right way.
1
0
4
@sarath_suresh_m
sarath menon
1 month
Backend development is over. Business logic is now natural language.
@sarath_suresh_m
sarath menon
1 month
Demo 1: Built a video fact-checker Nova in 20min Paste any YouTube URL → agent transcrisbes, extracts claims, verifies against sources, returns confidence scores. All business logic lives in the agent prompt. Zero backend complexity.
1
0
1