try this - https://github.com/ojuschugh1/sqz

I got tired of watching Codex re-read the same files over and over. A 2,000-token file read 5 times = 10,000 tokens gone. So I built sqz.

The key insight: most token waste isn't from verbose content - it's from repetition. sqz keeps a SHA-256 content cache. First read compresses normally. Every subsequent read of the same file returns a 13-token inline reference instead of the full content. The LLM still understands it.

Real numbers from my sessions:

File read 5x: 10,000 tokens → 1,400 tokens (86% saved)

JSON API response with nulls: 56% reduction (strips nulls, TOON-encodes)

Repeated log lines: 58% reduction (condenses duplicates)

Stack traces: 0% reduction (intentionally — error content is sacred)

That last point is the whole philosophy. Aggressive compression can save more tokens on paper, but if it strips context from your error messages or drops lines from your diffs, the LLM gives you worse answers and you end up spending more tokens fixing the mistakes. sqz compresses what's safe to compress and leaves critical content untouched. You save tokens without sacrificing result quality.

It works across 4 surfaces:

Shell hook (auto-compresses CLI output)

MCP server (compiled Rust, not Node)

Browser extension (Chrome + Firefox (currently in approval phase)— works on ChatGPT, Claude, Gemini, Grok, Perplexity)

IDE plugins (JetBrains, VS Code)

Single Rust binary. Zero telemetry. 549 tests + 57 property-based correctness proofs.

cargo install sqz-cli

sqz init

Track your savings:

sqz gain # ASCII chart of daily token savings

sqz stats # cumulative report Where sqz shines Scenario Savings Why

Repeated file reads (5x) 86% Dedup cache: 13-token ref after first read JSON API responses with nulls 7–56% Strip nulls + TOON encoding (varies by null density) Repeated log lines 58% Condense stage collapses duplicates Large JSON arrays 77% Array sampling + collapseToken Savingssqz saves tokens in two ways: compression (removing noise from content) and deduplication (replacing repeated reads with 13-token references). The dedup cache is where the biggest savings happen in real sessions.Where sqz shinesScenario Savings WhyRepeated file reads (5x) 86% Dedup cache: 13-token ref after first readJSON API responses with nulls 7–56% Strip nulls + TOON encoding (varies by null density)Repeated log lines 58% Condense stage collapses duplicatesLarge JSON arrays 77% Array sampling + collapseWhere sqz shinesScenarioSavingsWhyRepeated file reads (5x)86%Dedup cache: 13-token ref after first readJSON API responses with nulls7–56%Strip nulls + TOON encoding (varies by null density)Repeated log lines58%Condense stage collapses duplicatesLarge JSON arrays77%Array sampling + collapseToken Savingssqz saves tokens in two ways: compression (removing noise from content) and deduplication (replacing repeated reads with 13-token references). The dedup cache is where the biggest savings happen in real sessions.Where sqz shinesScenario Savings WhyRepeated file reads (5x) 86% Dedup cache: 13-token ref after first readJSON API responses with nulls 7–56% Strip nulls + TOON encoding (varies by null density)Repeated log lines 58% Condense stage collapses duplicatesLarge JSON arrays 77% Array sampling + collapse

GitHub: https://github.com/ojuschugh1/sqz

Happy to answer questions about the architecture or benchmarks. Hope this tool will Sqz your tokens and save your credits.

f you try it, a helps with discoverability — and bug reports are extra welcome since this is v0.2 so rough edges exist.

It is available as IDE Extension , CLI via npm, cargo, pip , soon it will be able as web extension to use with chatgpt, claude, gemmini websites as well.

I have been building on 3 tools lately all opensource-:

1) Etch-Detect API changes automatically using real traffic (no tests required)

https://github.com/ojuschugh1/etch

2) GhostDep-Detect phantom and unused dependencies across multiple languages using a fast Rust-based CLI

https://github.com/ojuschugh1/ghostdep

3) a local CLI that verifies whether AI coding agents actually did what they claimed

https://github.com/ojuschugh1/claimcheck

I have been building on 3 tools lately all opensource-:

1) Etch-Detect API changes automatically using real traffic (no tests required)

https://github.com/ojuschugh1/etch

2) GhostDep-Detect phantom and unused dependencies across multiple languages using a fast Rust-based CLI

https://github.com/ojuschugh1/ghostdep

3) a local CLI that verifies whether AI coding agents actually did what they claimed

https://github.com/ojuschugh1/claimcheck

I have been building on 3 tools lately all opensource-:

1) Etch-Detect API changes automatically using real traffic (no tests required)

https://github.com/ojuschugh1/etch

2) GhostDep-Detect phantom and unused dependencies across multiple languages using a fast Rust-based CLI

https://github.com/ojuschugh1/ghostdep

3) a local CLI that verifies whether AI coding agents actually did what they claimed

https://github.com/ojuschugh1/claimcheck

It is a rust-based cli. Looking for honest feedback on this :)

Thanks

Location: India

Remote: Yes

Willing to relocate: Prefer remote

Technologies: Open Source, Python, C/C++, LLMs, Gen AI, Rags, Git, Github, Docker, etc

Résumé/CV: https://www.linkedin.com/in/ojus-chugh-364328201/

I am looking for an entry-level software engineer position full-time.Past experience with Google, SAS Inc, Eutech Singapore

Location: Delhi, India

Remote: Yes

Willing to relocate: Only within the Delhi/NCR region, prefer remote

Technologies: C, Python, C++, Generative AI, HTML/CSS, Flask, Git, MERN, AI/ML, OpenCV, Open-Source, PHP, Java, CNN, SQL, Cloud, Software Development, Linux, LLMs, RAG

Résumé/CV: https://drive.google.com/file/d/1SZxHFXaXUKBKd4nOFq7ADKTbd4X...

Email: ojuschugh(at)gmail(dot)com

I am a recent IT graduate from India seeking full-time software engineer positions, preferably remote. I have experience working with LLMs, fine-tuning LLMs, RAG, open-source software development, machine learning, and web development. Recently, I worked at SAS as a Gen AI intern, I have done the Google Summer of Code at the Eclipse Foundation and worked as a full-stack intern at Eutech as well in the past. I am passionate and dedicated, with a current focus on Generative AI and LLMs.

I am only interested in full-time positions, not contract roles. However, I am open to research positions or internships that involve publishing research papers. Feel free to reach out to me.

Thank you.

Location: Delhi, India

Remote: Yes

Willing to relocate: Yes but prefer remote

Technologies: Java, J2EE, Python, C++, TypeScript, JavaScript, React.js, Next.js, Node.js

Résumé/CV: [Link to Résumé/CV](https://drive.google.com/file/d/1gi4ve8rPmS_IWAQwU07bCmvvt1C...)

Email: ojuschugh@gmail.com

GitHub: [GitHub Profile](https://github.com/ojuschugh1/)

Happy New Year Everyone! I hope you all are doing well.I am Ojus, a recent IT graduate and I'm looking for a job. I can join immediately. I'm interested in Software Development, C++,Python, Java EE, Jenkins, Debugging, AWS, Open Source, Web Development, and I'm open to trying out something new.

I have worked as a software engineering intern at Eclipse Foundation (Founded by IBM) as part of Google Summer of Code 2023, working on Eclipse Cargo Tracker with Reza Rahman (principal program manager at Microsoft).

Most of my programming experience during college has been with C++, Javascript and Python, and for the past four months, I’ve been writing code in Java by working on Eclipse Cargo Tracker, upgrading the codebase from Jakarta EE 8 to Jakarta EE 10 (Formerly Java EE) as part of Google Summer of Code. Read more about my GSoC contributions: [Link to Medium Article](https://medium.com/@ojuschugh1/gsoc23-work-product-submissio......)

Apart from that, I'm also interested in DevOps and Gen AI (still in the early stages of learning it).

I am committed to proving my capabilities and dedication if given the chance. Thank you for considering my request.

Thanks, Ojus.

Hi,Thanks for your help. I wanted to post the comment on the ask HN:who wants to hired but unbale to comment so i was figuring out over internet but unable to find it.

Right now as well I am unable to comment in the ASK:HN who wants to be hired(December,2023)even after you vouch me, why so ?.

You posted in https://news.ycombinator.com/item?id=38666225 that is the wrong thread. Are you trying to repost that in the correct thread? Perhaps the system detects the duplication??? (I never hear that, but there are a few hidden features.)

Try again, and if that fails, send an email to the mods hn@ycombinator.com

ok , could you kindly please tell like how to comment or post in a thread may be i am doing in the wrong way like I post through submit option , in title i add some random title related to text , and i keep url blank and write the description or text in the text box and then submit it.Is this i am doing in the right way ?.

For ask section thread, i can see option to reply to others comment but unable to comment my own message.

Thanks and Regards

You posted in "Ask HN: Who is hiring? (December 2023)" https://news.ycombinator.com/item?id=38490811 that is for business that want to hire a person

You should have posted in https://news.ycombinator.com/item?id=38490809 "Ask HN: Who wants to be hired? (December 2023)" that is for persons that want to be hired by a business.

I'm just looked at my calendar, and it's already 18th. I think the threads close automatically after some time (15 days?). So both may be closed. I think you should wait until January 2nd and post in the new thread.

Thanks for letting me know about this, I recently got confirmation that they are closed.

Thanks to both you for the help :)