Shaan

20 | Engineer | CS | Open to SWE Intern and SWE-1 roles

[Shaan @shaan_dv]

@BuildACrewHQ DONE

[Shaan @shaan_dv]

@imharshhub https://t.co/jWY2JrXbZN

[Shaan @shaan_dv]

@iradotso Is any ios invite left?

[Shaan @shaan_dv]

@caucasianx_99 @X Done

[Shaan @shaan_dv]

Hey @X algorithm ☺️

I'm looking to #connect with people interested in:

Full-stack
Frontend
Backend
Next.js / React
TypeScript
Prisma & Databases
Open Source
Software Engineering
Web Dev
SaaS & Startups

Drop what you're working on below! Lets connect 🤝

[Shaan @shaan_dv]

@wespreadjam audit

[Shaan @shaan_dv]

gta7 launched before gta6

[Sundar Pichai @sundarpichai]

Excited to share Genie 4 today our most realistic and interactive world model:

- steer dynamic environments in real time, shaping events, camera angles, and actions

- generate native audio matched to the world, from realistic background sounds to characters who can converse in real time!

- simulate photorealistic scenes and real-world physics

- create more coherent multi-step events through reasoning from Gemini

We've started to roll out as a research preview to trusted testers. Coming soon to Google AI Ultra subscribers in Project Genie and early access partners via the Gemini API and AI Studio.

[Shaan @shaan_dv]

Honest limits: TypeScript + Jest only, tested on a demo repo with planted bugs. Next: mutation testing and more languages.

Code: https://t.co/1us68ntrSC

[Shaan @shaan_dv]

I built TestPilot at the Alan hackathon: an AI agent that reads a code change, figures out what it could break, writes a regression test, and proves the test catches the bug. Then it proposes a fix. 🧵 @prasad_pilla @tryAlanAI

[Prasad Pilla @prasad_pilla]

Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.

We kept running into the same problem ourselves.

Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.

We built Alan to change that.

[Shaan @shaan_dv]

Built with Alan from planning to the actual build and open models mainly GLM 5.3 and deepseek 4.1 flash. The engine uses the Vercel AI SDK to swap between Claude, OpenAI and Gemini, and TestPilot ships as an MCP server + skill so Alan can call it.

[Shaan @shaan_dv]

How it works:

Maps the diff to callers and untested paths
Generates a Jest test for the risk
Runs it on base AND changed code: pass on base, fail on the change = evidence
Proposes a fix, re-verifies

Results come from the test runner, not the LLM.

[Shaan @shaan_dv]

The problem: coding agents ship changes faster than anyone can test them.

A “harmless” perf tweak (Promise.all in a payment flow, .lean() on a user lookup) passes every existing test and still breaks behavior.

Green CI isn’t proof

[Shaan @shaan_dv]

AI can write code in seconds. But can it tell you what that code might break?

Building TestPilot, an AI agent that analyzes code changes, identifies regression risks, generates targeted tests, and verifies they actually catch bugs.

@prasad_pilla @tryAlanAI

[Shaan @shaan_dv]

@notanilp SAPLING

[Shaan @shaan_dv]

@ayushGup7 badminton/cricket & poker

[Shaan @shaan_dv]

@boardyai @hf0 30

[Shaan @shaan_dv]

@iamanishx @ujsquared Yes, I think 1 week paid trials are a good option for startups

[Shaan @shaan_dv]

@iamanishx @ujsquared There are even more ways to cheat like cluely obv, even OS level software is there which cant be caught even if you give all the access to a safe browser sort of product, they bind themselves with the process of the browser, so they dont get detected as an external app

[Shaan @shaan_dv]

@_Shiva_iitp https://t.co/ORQ3w9H325

[Shaan @shaan_dv]

@ClaudesRH hello

[Shaan @shaan_dv]

@ayushkeshri4k This is old