ticktockten
Show HN: Fast-Axolotl – Rust extensions that make Axolotl fine-tuning 77x faster
https://github.com/neul-labs/fast-axolotl
I built Rust extensions for Axolotl that dramatically speed up data loading and preprocessing for LLM fine-tuning.The problem: Python data pipelines become the bottleneck when fine-tuning large models. Your GPUs sit idle waiting for data.The solution: Drop-in Rust acceleration. One import line, zero config changes.Results on 50k rows: - Streaming data loading: 0.009s vs 0.724s (77x faster) -...
Show HN: SolScript – Write Solidity, compile to Solana programs
https://github.com/cryptuon/solscript
Hey HN, I built SolScript, a compiler that lets you write smart contracts in Solidity syntax and deploy them to Solana.The problem: Solana has mass dev interest (17k+ active developers in 2025), but the Rust learning curve remains a 3-6 month barrier. Anchor helps, but you still need to grok ownership, lifetimes, and borrowing. Meanwhile, there are 30k+ Solidity developers who already know how...
Show HN: SolScript – Write Solidity, compile to Solana programs
Hey HN,I built SolScript, a compiler that lets you write smart contracts in Solidity syntax and deploy them to Solana.The problem: Solana has mass dev interest (17k+ active developers in 2025), but the Rust learning curve remains a 3-6 month barrier. Anchor helps, but you still need to grok ownership, lifetimes, and borrowing. Meanwhile, there are 30k+ Solidity developers who already know how...
Re:BJH OS – Free Open Source Browser OS for Windows, Mac, Linux, Andro...
on Wednesday December 10, 2025 @06:50AM (#46216697)
Looks cool :).
What does BJH stand for though?
Show HN: LangGraph profiling – 737x Faster Checkpoints via Rust (PyO3)
https://github.com/neul-labs/fast-langgraph
Building AI agents with LangGraph, I noticed graph invocations were slow even before hitting the LLM. Dug into the Pregel execution engine to find out why.THE PROBLEMProfiled my LangGraph agents. 50-100ms per invocation, most of it not the LLM. Found two culprits:1. ThreadPoolExecutor created fresh every invoke() — 20ms overhead2. Checkpointing uses deepcopy() — 52ms for 35KB state, 206ms for...
Re:Show HN: Optimizing LiteLLM with Rust – When Expectations Meet Reality
on Tuesday November 18, 2025 @02:38PM (#45970978)
Well several times faster, but not interesting enough to say that use this. For me it personally was an exploratory project to review litellm and its internals.
The LLM docgen in this case Claude has been over enthusiastic due to my incessant prodding :D.
Re:Show HN: Optimizing LiteLLM with Rust – When Expectations Meet Reality
on Tuesday November 18, 2025 @02:33PM (#45970899)
Well i would counter that by saying most code has been autocompleted for a while. At this point in software development history, discussing the size of commits is a null discussion :).
Re:Show HN: Optimizing LiteLLM with Rust – When Expectations Meet Reality
on Tuesday November 18, 2025 @02:31PM (#45970873)
The core is real, the rest of the narrative nudging LLMs to behave :). If you remove the noise and just run the benchmark that's proof enough.
The interesting bit was that the bindings overheads dominate, and makes this shim not that much of a performance bump.
Show HN: Optimizing LiteLLM with Rust – When Expectations Meet Reality
https://github.com/neul-labs/fast-litellm
I've been working on Fast LiteLLM - a Rust acceleration layer for the popular LiteLLM library - and I had some interesting learnings that might resonate with other developers trying to squeeze performance out of existing systems.My assumption was that LiteLLM, being a Python library, would have plenty of low-hanging fruit for optimization. I set out to create a Rust layer using PyO3 to...
Show HN: FastWorker – Task queue for Python with no external dependencies
https://github.com/neul-labs/fastworker
I built FastWorker after getting tired of deploying Celery + Redis for simple background tasks in FastAPI apps. Every time I needed to offload work from API requests, I had to manage 4-6 separate services. For small projects, this felt like overkill.FastWorker is a brokerless task queue requiring only Python processes. No Redis, no RabbitMQ – just 2-3 Python services instead of 4-6+.---Quick...
Re:Show HN: Zyler – AI agent for marketing data that doesn't hallucinate
on Tuesday June 24, 2025 @07:52AM (#44365155)
Yes! We are looking to add Meta, Linkedin and titok. There is a lot of demand for cross channel analytics for them.
Re:Show HN: Zyler – AI agent for marketing data that doesn't hallucinate
on Tuesday June 24, 2025 @06:39AM (#44364698)
Thanks for checking us out!
We are looking to integrate more data connectors.
Do you have any suggestions? We are focused on the google ecosystem to begin with.
Show HN: Zyler – AI agent for marketing data that doesn't hallucinate
My co-founder Suryansh (product marketing expert) and I were burning way too many nights debugging why our startup's conversion rates were tanking. I'd spent years building adtech and martech software, but we were still juggling between Google Analytics, Google Ads, YouTube Analytics, and SEO tools – each with their own dashboard that felt like archaeological expeditions. After burning 6+ hours...
Re:Show HN: Unzoi – Research tool that minimizes information misrepres...
on Tuesday April 01, 2025 @05:57AM (#43544878)
Hey, can you try again?
Thanks!
Re:Show HN: Unlob – A search engine that reads websites and gives you ...
on Monday March 31, 2025 @02:40PM (#43538247)
That makes sense, specifically for technical guides.
It not well designed for those cases for sure.
Would you be up for a chat if i drop you an email? More than happy to learn your general views :)
Re:Ask HN: What are you working on? (March 2025)
on Monday March 31, 2025 @11:29AM (#43536171)
I have been working towards building and understanding the shift in search / information retrieval.
To that end, I just did a show HN on a couple of my projects
What if google did not focus on hyper commercialisation? 10 Blue links, but sorted by how they answer your query- https://www.unlob.com
Can we answer questions with lesser hallucinations? A snippet cited answer engine which only picks links that focus on answering your query - https://www.unzoi.com
Re:Show HN: Unlob – A search engine that reads websites and gives you ...
on Monday March 31, 2025 @11:25AM (#43536119)
Here is a followup from the other tool i am working on
https://www.unzoi.com/query/set-up-model-context-protocol-se...
Show HN: Unzoi – Research tool that minimizes information misrepresentation
I built Unzoi, a research tool designed to address critical issues I've observed with existing AI assistants and search tools.Current research tools often struggle with accurately representing source material. They frequently introduce factual errors, misattribute quotes, present opinions as facts, and fail to provide sufficient context.Many AI assistants also add their own interpretations or...
Re:Show HN: Unlob – A search engine that reads websites and gives you ...
on Monday March 31, 2025 @06:59AM (#43533501)
Well the goal is not to take anyone's job. Ideally, we make our present tools a lot better and trustworthy.
Thanks for trying it out!
Show HN: Unlob – A search engine that reads websites and gives you direct answer
I built Unlob because I was tired of search engines that only show links and AI tools that hallucinate answers. Unlob actually reads websites for you and extracts specific information directly from reliable sources.Instead of: - Wading through pages of SEO-optimized garbage- Getting vague AI-generated text with no citations- Opening 10+ tabs to find one piece of informationUnlob gives you...
Show HN: GenAI Playbook – Open-source guide to implement GenAI in business
https://www.whatgenerativeai.com/
I've created an open-source playbook to help businesses implement Generative AI effectively. It covers:- Identifying high-impact AI use cases - Building and integrating AI models - Navigating ethical and regulatory challenges - Fostering an AI-driven cultureThe playbook is based on my 15+ years in AI and blockchain, including work with LLMs and computer vision.Roadmap: - GenAI Cookbook with...
Show HN: Fragaria – From 'R's in Strawberry to Complex Problem-Solving AI
https://github.com/terraprompt/fragaria
I'm excited to share Fragaria, an open-source project that brings advanced AI reasoning capabilities to developers and researchers. Named after the botanical genus of strawberries (a nod to the classic "How many 'r's in strawberry?" problem), Fragaria is designed to tackle both simple queries and complex logical puzzles with increasing efficiency over time. Here are some features!*...
Re:Drive drunk twice in Quebec, get ignition breathalyzer for life
on Monday December 02, 2019 @12:50AM (#21680131)
This should be standard issue for cabs!
Re:Instagram’s Explore Recommender System
on Tuesday November 26, 2019 @02:17PM (#21641607)
Yeah, i have come to a similar conclusion. It is also very clear that academic papers are unless simple enough to be understood (like word2vec) they rarely make it to production.
Re:Instagram’s Explore Recommender System
on Tuesday November 26, 2019 @02:09PM (#21641540)
Interesting, and looks like the present choices have evolved over time.
Seems like they are moving towards a structured RL implementation. There are elements of it, a follow-up post on some components would be interesting.
Re:Show HN: Coscreen.co – a radically different remote collaboration tool
on Tuesday November 26, 2019 @01:57PM (#21641423)
Look forward to using this in linux! Definitely a tool I would use.
TF-Agents: A Flexible Reinforcement Learning Library for TensorFlow
AWS's Deepracer to Teach Everyone Reinforcement Learning
Re:Cthulhu Ventures
on Tuesday November 26, 2019 @04:51AM (#21636867)
I had no idea! Thanks for sharing.
Having spent so much time with flask, it's a joke that I took seriously enough :)