Adding Java class inline editing without decompilation feature to https://jar.tools/class-decompiler - so you can update some class values (number or strings, or event access modifiers) without needing to recompile the .class file. I called it Tweaks

MacroCodex app, which results in guaranteed weight loss or weight gain outcomes within 2-5 weeks. It's a completely ad-free, subscription-free product.

Don't believe? Read the reviews.

All you need to do is provide it your calorie intake and weight reading each day.

How does it work? Explained here: https://macrocodex.app/knowledge/rethink/adaptive-tdee/

We already have 16,000+ users.

A visually grounded AI tutoring application: https://knowable.ca Been working full-time on it for a few months, I'm trying to make the concept of an AI tutor better than just a chatbox.

I wrote a release train modelling app (https://allaboard.ooer.com/) and it turned out so horribly complex and hard to use I'm building a general systems modelling app instead so I can rebuild the model in a nice text format.

I've been working on two projects.

## Tab Highway: Browser extension focusing on the UX/process of decluttering

It turns tab cleanup into a game.

When you pile up too many open tabs or whenever you want, it opens a night drive where each tab is a road sign you drive through or send to the trash.

[GitHub](https://github.com/pravj/TAB-HIGHWAY)

This is still a work in progress. It works on my machine, but not fully distribution-ready.

## A short satirical blog on the overly-expressive/interactive web experiments fueled by coding agents.

Ironically, the project listed above falls in the same category.

[Woodworking Has Been Achieved Internally](https://hackpravj.com/blog/woodworking-has-been-achieved-int...)

Work: Security engineering focused on securing authnz at hyperscale in our new agentic reality. What interests me? Agentic engineering, vulnerability discovery at massive scale. Vulnerability remediation at massive scale. Bounding agent non-determinism.

Personal: Local models. Agent harnesses. Track cars, turning wrenches.

Had agents build recently:

- Photo catalog app with all 100K photos in my library described by a local 9B parameter model (it is absolutely amazing and my favorite personal project in terms of genuien impact)

- Generic RAG layer (vector embedding + FTS in SQLite with RRF and document ingestiation layers for 1-2GB knowledge bases to power agent context for stuff I am interested in, neuroscience and philosophy is where I spend time on this right now, and basically every AI research paper publicly available to help agents tinker with local models). This is especially useful for local models, but it even helps frontier models because they can freely churn through without a bunch of web calls and such when you are on a topic thread or research.

- Qwen 3.6 35B A3B clean-up proxy, achieving a clean thinking + tool calling A3B that I can leave running for 24+ hours. This particular model gets stuck in long reasoning chains and needs encouragement to get on with it, especially on context tasks. It actually reasons well but then starts second guessing and loops. Built a deterministic framework to inspect its thinking and make it wrap its thinking up if it gets stuck and also to trigger calls to external models and inject solutions to problems when it gets stuck. The agent sees the solution injected from a remote model like it is its own thinking, kind of a subconscious it has no idea about. Metacognition in models, it is wild.

DSV4->GLM->Kimi is the "phone a friend" external brain pathing I built. As far as I know no one has shared all of the bugs I have 'fixed' in my project and gotten local Qwen A3B this reliable.

I've been hacking away on some scripts to clean up saved ChatGPT transcripts (surprisingly annoying because of how the app's JavaScript works, and I discovered too late that some of them were incomplete; but I have workarounds now) and convert them to a fairly minimal Markdown with some custom restructuring. This is mainly so I can share them on my blog and discuss some experiences of using the LLM; but also I'm extracting some code samples as the basis of some small projects I've had in mind for years but could never motivate myself to start (at least partly from an inability to choose).

In particular: an image viewer with a small suite of extra functionality; and a browser extension for cleaning up URLs and then bookmarking them in a more organized way that I might actually use meaningfully.

A modern issue tracker enabling replay for agentic workflows (via board time-travel) and focus on ergonomics (native to the terminal).

https://ljtn.github.io/epiq/

I've finalized and have in beta testing an AI-based natural language search of properties from an MLS feed. On top of that, I'm continually tuning an AI agent that is able to do property research, comparisons, neighborhood and history analysis, etc.

[1]: https://mlsync.ai

I resurrected an ancient UIUC x86 ASM game, built with some friends in the summer of 2002 for ECE291. Claude Code drove most of the work.

The first attempt was a full on x86asm->webasm transpiler which ran into some difficult stack issues.

The second attempt was successful via conversion from x86asm->c->emscripten->wasm [0]. It plays pretty well on mobile and just about any browser now.

Fair warning it is pretty difficult to win, unless you can figure out where the enemy weaknesses are [1].

[0] [code] https://github.com/gottebp/alan_parsons_project

[1] [play] https://particlefield.com/projects/alan-parsons/game.html