HNHacker News
TopNewBestAskShowJobs

srijanshukla18

24 karma · joined November 22, 2016

submissionscomments
srijanshukla18··on Ask HN: How do you manage skills files?
I've got a skill repository on github, and I got a hermes automation to sync skills repo - this hermes automation is on every device. Any skills I make on the fly - my global AGENTS.md has instructions to update the skills repo path and place them appropriately in there.
srijanshukla18··on Show HN: Spin Lab
Aesthetic fatigue from AI :D I understand, I feel it too.

But yeah I've got zero UI skills - I just wanted to showcase the spin dynamics somehow, so shared it anyway.

srijanshukla18··on Show HN: Spin Lab
Yeah sorry about that. Doesn't look that good on the phone. Try it out once on your laptop/desktop - much better there.

And actually it should all work - just looks cramped on phone.

srijanshukla18··on Show HN: Spin Lab
Got it, I can edit that bit, thanks for the feedback.
srijanshukla18··on Show HN: Spin Lab
Yes! it does do the math for Magnus effect. Source code is all client side you can take a look at it.

But early on when I started this, I prompted the LLMs strongly to focus on having accurate real world physics baked in.

srijanshukla18··on Show HN: Spin Lab
Fair feedback. I've got zero UI skills - I just did put up whatever Fable gave out. That's true.

I put tons of time explaining spin physics to Fable - that was some effort on my part and that's it.

the UI feedback, I'll pass it Fable when its back :D

And yep it doesn't look good on mobile devices - try it on a desktop screen if you can, much better there.

srijanshukla18··on Show HN: Spin Lab
Yeah no idea but I've seen devices that are able to measure spin in real time.
srijanshukla18··on Show HN: Spin Lab
For sure yeah just scratches the surface for now

About the bounce and the answer: great feedback, thank you! When fable is back I'll incorporate these.

But no did not take just 5 minutes :D I was working on this since early Opus days and the output was consistently absolutely horrible. Fable was the first model to produce something okay - then I prompted it for hours explaining exactly what I want to demonstrate and had to even tell it explicitly a lot about how spin works - from my experience

srijanshukla18··on Show HN: Spin Lab
Hey HN, I built Spin Lab: a browser-based interactive explainer for table-tennis spin.

It visualizes topspin/backspin, spin rate, ball trajectory, bounce behavior, and why the opponent’s return reacts the way it does.

I built it because spin is central to table tennis, but most explanations are either too hand-wavy or too static. I wanted something you can scrub, slow down, and experiment with.

srijanshukla18··on ProgramBench: Can language models rebuild programs from scratch?
This is not a serious benchmark, come on.

Tomorrow I'm launching a benchmark where I check if an LLM can build a Airbus A320 from scratch without internet. (Spoiler: no LLM succeeds)

srijanshukla18··on Benchmarking OpenTelemetry: Can AI trace your failed login?
Humans can't do much OTelBench Try finding even good documentation for it

That's just misleading phrasing on this post

I'm an SRE, AI does NOT struggle with 'simple SRE tasks' OTel instrumentation by no measure is a 'simple SRE task'

srijanshukla18··on Getting a Gemini API key is an exercise in frustration
Yes, tried countless hours and even reached out to google billing support. Doesn't accept UPI or card, runs into an error with both.
srijanshukla18··on Show HN: Avoid LLM Sycophantry – Two Models Debate Any Claim
I actually had the exact same idea and wanted to build it! Good on you thanks for building this.
srijanshukla18··on Show HN: XRAY MCP – AST-grep wrapped in a tiny server for code-aware AI
hey thanks for checking it out!
srijanshukla18··on Show HN: XRAY MCP – AST-grep wrapped in a tiny server for code-aware AI
Hey, thanks for checking it out!
srijanshukla18··on Show HN: XRAY MCP – AST-grep wrapped in a tiny server for code-aware AI
exactly yep, hate it when it just keeps going in circles. thats a lot of extra tokens also you know! let me know your thoughts once you check it out
srijanshukla18··on Show HN: XRAY MCP – AST-grep wrapped in a tiny server for code-aware AI
You've hit the nail on the head! "Understanding the blast radius" is the core problem xray mcp tries to solve.

The whole idea behind tools like what_breaks is to give the agent that foresight before it makes a change, moving it from a reactive "change -> lint -> fix" cycle to a more proactive "plan -> analyze -> change" workflow.

Thanks for checking it out!

About the linting use case specifically, I've found claude code hooks to be very beneficial. Not sure if other tools have an equivalent. Check that out too!

srijanshukla18··on Show HN: XRAY MCP – AST-grep wrapped in a tiny server for code-aware AI
I know right! I am surprised that many of these cli tools just keep running around the codebase read whole files at once and grepping vaguely
srijanshukla18··on Show HN: XRAY MCP – AST-grep wrapped in a tiny server for code-aware AI
Yes! depends on the code agent you are using. CLAUDE.md or GEMINI.md for example - you can be adding a one liner there - "always use xray mcp tools to make any changes or to explore a new codebase"
srijanshukla18··on Show HN: FFlags – Feature flags as code, served from the edge
Great product! Actually a big point that I've seen in enterprises.
srijanshukla18··on Show HN: Hybrid Knowledge Graph and RAG for Legal Documents (Learning Project)
oh great question, hadn't thought of that. But this particular project the TF-IDF vectors and KG building is all local, and the main.py builds both during every start can be adding a hook to listen for file changes and rebuild KG and vectors I guess