
Building https://xitter.gq/t.co/P0putPtdEX & https://xitter.gq/t.co/m962rg1TBc | He/Him | Staff Engineer (Mobile) - @MutualMobile | Thoughts are my own
Hyderabad, India
Joined March 2010
- Tweets1.3K
- Following717
- Followers255
- Likes42.4K
Shreyansh Lodha retweeted
Replying to @Google
@Google recently announced AppFunctions in Google I/O which let's the an Android device's local AI agents to trigger and access your app's functionalities.
So far we have focussed on making our apps usable to humans now it's time to make them usable for AI Agents.
github.com/avidraghav/AgentN…
Shreyansh Lodha retweeted
🚀 Introducing Debroid — the headless Android debugger built for AI coding agents.
AI agents can write Android code… but they’re blind at runtime.
Debroid gives them eyes and hands for debugging:
🔴 Set breakpoints
⚡ Catch exceptions
👀 Watch fields
🔬 Inspect & mutate live memory
🐾 Step through code
🧮 Evaluate expressions
All via CLI → No GUI needed.
Works with your favorite agents😉.
Install: github.com/PatilShreyas/debr…
Give your agent real runtime superpowers 🤖⚡
Hi everyone 👋
I’m exploring new Senior / Staff / Principal Android engineering opportunities. If you know of a role that could be a fit, I’d really appreciate an intro or a quick chat (DMs are open).
A little about me:
• 14+ years in native Android across Equal AI, JioCinema (Viacom18), Gojek, Meesho, and DoorDash - products used by 100M+ people
• Google Developer Expert for Android since 2021. Built Android from zero to 450K DAU at Equal AI, led the Core Mobile Platform team at JioCinema through IPL-scale traffic, and shipped weekly KMP libraries powering Android, iOS, and Web
•Deep in Kotlin, Compose, KMP, and mobile platform engineering - plus agentic AI tooling: MCP servers, spec-driven development, and open-source work like compose-skill (480+ ⭐), Andromeda, and ComposeProoof
•Comfortable owning technical direction end to end, from architecture strategy to production, while staying hands-on in the code
I’m open to remote contract work (worldwide) and permanent Staff/Principal roles in Europe with relocation support.
P.S. If your intro leads to my next role, I’ll gift you lifetime access to the agentic mobile tooling I’m building.
I build Android that doesn't page you at 3am.
14 yrs, GDE. 0 → 450K DAU on my last app; KMP at 100M+ scale before that. Lately building agent tooling for Compose (ComposeProof.dev)
and
github.com/aldefy/compose-sk…
github.com/aldefy/compose-re…
Taking on a remote contract or two Principal/Staff.
Hiring, or know someone who is? Tag them 👇
Shreyansh Lodha retweeted
Replying to @bcherny
I have to push back on 2 things as i think one is categorically incorrect and the other is demonstrably incorrect.
1. Debugging:
Debugging is not a thing if coding is solved. You would produce correct behaviors. I don't understand how a solved problem could produce erroneous behavior.
2. Coding is the easy part:
setting hardware, capacity, talking to users, product planning agreed is in fact hard, but so is coding.
Example: If coding was in fact not hard then Claude Code having a flickering issue for well over 9 months, which is a purely software challenge, would have been solved almost immediately (immediately being on a shortened time scale comparatively to a human solve time scale).
For more trivial applications software approximation can largely work. I also love software approximation for exploring how things should feel.
Shreyansh Lodha retweeted
I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem.
As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)!
I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work.
It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results?
88ms => 1.5ms
150K allocs => ~500 allocs
Incredible right? Nope.
My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path.
This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput.
The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity.
Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.
Shreyansh Lodha retweeted
5 Minute Bedtime Stories For Android Devs.
This book, brought together by James Cullimore and co-authored by 30 renowned developers, was one of the most interesting books I’ve read recently.
I also joined the "An Open-Source Nightmare at Midnight" section.
Shreyansh Lodha retweeted
When I hear someone say "AI is like a senior developer" all I can think of is you and I have a very different defintion of the "senior developer".
It will help to learn to leverage the tools and spare the comparisons to humans' experience or capabilities.
Shreyansh Lodha retweeted
Imagine being the president of Y Combinator and blocking every good engineer on this app because they pointed out you were using the wrong metric and instilling future generations with bad advice.
Shreyansh Lodha retweeted
Oh, it just got worse. The [public github issue](github.com/BerriAI/litellm/i…) has been closed as "not planned" by the owner, so they likely have been fully compromised.
Shreyansh Lodha retweeted
She literally explained that the discomfort of discipline is nothing compared to the pain of regret.
this guy has 29 models on huggingface at page 2 ranking. no lab behind him. no sponsorship. $2,000 from his own pocket on GPU rentals. he compressed GLM-4.7 to run on a MacBook and quantized Nemotron Super the week it dropped. all public. all free.
nvidia is a trillion dollar company with hundreds of teams but they are not the ones quantizing models middle of the night and pushing them out before sunrise. if nvidia stopped tomorrow their employees stop working. people like @0xSero would not. that is the difference between a paycheck and a mission.
@NVIDIAAI you talk about making AI accessible. the people actually doing it are right here. 29 models deep burning their own compute with no ask except more hardware to keep going. you do not need to build another program. just look at who is already building for you. one GPU to this man would produce more public value than a hundred internal sprints.
i am not asking for charity. i am asking you to invest in someone who already proved it.
Putting out a wish to the universe.
I need more compute, if I can get more I will make sure every machine from a small phone to a bootstrapped RTX 3090 node can run frontier intelligence fast with minimal intelligence loss.
I have hit page 2 of huggingface, released 3 model family compressions and got GLM-4.7 on a MacBook huggingface.co/0xsero
My beast just isn’t enough and I already spent 2k usd on renting GPUs on top of credits provided by Prime intellect and Hotaisle.
———
If you believe in what I do help me get this to Nvidia, maybe they will bless me with the pewter to keep making local AI more accessible 🙏




























