Aakash

A builder of 10+ years turned founder. Open source FTW!

Co-Founder of with me at

[Aakash @aakashH242]

@fil_ships Looking forward to your feedback. Happy to improve it based on your use-case. 😃

[Aakash @aakashH242]

@fil_ships If it helps, you can check out a library I wrote specifically for this purpose - trimming context smartly instead of truncation.

https://t.co/hTMpEq7gWc

[Aakash @aakashH242]

@dezelin Ah got it, the branch shortcuts were the bit I was curious about. Thanks for explaining!

[Aakash @aakashH242]

@fil_ships What do you do if a ticket starts pushing past 150k? Split it further, or trim the context and keep going?

[Aakash @aakashH242]

@maheshboda006 Good to see a side-by-side test. The save step would be interesting to dig into: can annotations made in either build be reopened and edited in Preview? That'd say a lot about file compatibility.

[Aakash @aakashH242]

@RABSWFL Nice to see more dictation tools on Linux! Can you skip Claude's cleanup for a single dictation? Keeping filenames and command flags verbatim seems especially useful when talking through code.

[Aakash @aakashH242]

@bravish_ Congrats on 1.0! The signed gaps caught my eye. A small replay showing a complete run alongside one with missing events would be a useful demo of what the verifier can and can't tell you.

[Aakash @aakashH242]

What's the smallest dev tool you've built that you still use every day?

A shell script counts. What annoying little problem did it solve?

[Aakash @aakashH242]

@KI_Vater Hey Maurice! I'm building open-source tools for AI coding too, including Gauntlet for code review. With Plugin Guard, what's been harder so far: catching risky behavior or keeping false positives manageable?

[Aakash @aakashH242]

@b4dgerr AI behind AI detector is hallucinating. :/

[Aakash @aakashH242]

@zmickyzee1 @ArsenalN7 He's fine. Played during the second half.

[Aakash @aakashH242]

@zxlzr Yeah, a useful test would be clearing the original task, then seeing whether the agent can recover a definition it needs later without being told to look for it.

[Aakash @aakashH242]

I just pledged to keep humans in control of AI. Join me. #TeamHuman https://t.co/w6X0aXNrsd

[Aakash @aakashH242]

@itzKashan2912 Try https://t.co/bLnt6q8P5o in a browser while logged into the affected account. Briefly explain why you think it was flagged by mistake. What happens when you try: an error, a disabled form, or trouble logging in?

[Aakash @aakashH242]

One rule in Gauntlet: a review pass that changes the code can’t also count as the clean pass.

Run the relevant checks on the final version, then give the repair a fresh review. A passing check only tells you about the version it ran against.

[Aakash @aakashH242]

@bigcodegabo Nice to see this built for Ubuntu. The tool cards showing what Neptune actually executed caught my eye in the docs. That should make it easier to check the lead agent’s summary when something goes wrong.

[Aakash @aakashH242]

@omarvvvr For a bug fix, I’d start with a reproducer that fails before the patch and passes after it. Then run the wider tests and try a nearby edge case. That helps check whether the fix addresses the actual bug.

[Aakash @aakashH242]

@b4dgerr Fair point 😅 Are you looking for a coding agent to use day to day, or a framework for building your own agents? That would help narrow the shortlist.