LumoSQL (Score: 151+ in 4 hours)
Link: https://readhacker.news/s/6vaim
Comments: https://readhacker.news/c/6vaim
Link: https://readhacker.news/s/6vaim
Comments: https://readhacker.news/c/6vaim
How a hawk learned to use traffic signals to hunt more successfully (🔥 Score: 152+ in 3 hours)
Link: https://readhacker.news/s/6vaqx
Comments: https://readhacker.news/c/6vaqx
Link: https://readhacker.news/s/6vaqx
Comments: https://readhacker.news/c/6vaqx
Frontiers Science News
Street smarts: how a hawk learned to use traffic signals to hunt more successfully
Dr Vladimir Dinets, a zoologist who studies animal behavior, ecology, and conservation, is the author of a recently published Frontiers in Ethology article that
BGP handling bug causes widespread internet routing instability (Score: 151+ in 5 hours)
Link: https://readhacker.news/s/6vamw
Comments: https://readhacker.news/c/6vamw
Link: https://readhacker.news/s/6vamw
Comments: https://readhacker.news/c/6vamw
blog.benjojo.co.uk
BGP handling bug causes widespread internet routing instability
Square Theory (🔥 Score: 156+ in 2 hours)
Link: https://readhacker.news/s/6vb3Q
Comments: https://readhacker.news/c/6vb3Q
Link: https://readhacker.news/s/6vb3Q
Comments: https://readhacker.news/c/6vb3Q
Adam Aaronson
Square Theory | Adam Aaronson
The story starts in Crosscord, the crossword Discord server. Over 5,000 users strong, the server has emerged as a central hub for the online crossword community, a buzzing, sometimes overwhelming, sometimes delightful town square where total noobs, veteran…
FromSoft's singular mech game Chromehounds is back online (Score: 150+ in 17 hours)
Link: https://readhacker.news/s/6v9tn
Comments: https://readhacker.news/c/6v9tn
Link: https://readhacker.news/s/6v9tn
Comments: https://readhacker.news/c/6v9tn
Read Only Memo
Interview: 15 years after the servers shut down, FromSoft's singular mech game Chromehounds is back online
The PvP mech battler is finally playable online in Xbox 360 emulator Xenia, and I talked to the modder making it happen.
DuckLake is an integrated data lake and catalog format (Score: 152+ in 5 hours)
Link: https://readhacker.news/s/6vaHQ
Comments: https://readhacker.news/c/6vaHQ
Link: https://readhacker.news/s/6vaHQ
Comments: https://readhacker.news/c/6vaHQ
DuckLake
DuckLake is an integrated data lake and catalog format.
DuckLake delivers advanced data lake features without traditional lakehouse complexity by using Parquet files and your SQL database. It's an open, standalone format from the DuckDB team.
Pyrefly vs. Ty: Comparing Python's two new Rust-based type checkers (Score: 153+ in 4 hours)
Link: https://readhacker.news/s/6vaWH
Comments: https://readhacker.news/c/6vaWH
Link: https://readhacker.news/s/6vaWH
Comments: https://readhacker.news/c/6vaWH
Edward Li's Blog
Pyrefly vs. ty: Comparing Python’s Two New Rust-Based Type Checkers
A deep dive into Meta's pyrefly and Astral's ty - two new Rust-based Python type checkers that both promise faster performance and better type inference.
I salvaged $6k of luxury items discarded by Duke students (Score: 150+ in 6 hours)
Link: https://readhacker.news/s/6vb8z
Comments: https://readhacker.news/c/6vb8z
Link: https://readhacker.news/s/6vb8z
Comments: https://readhacker.news/c/6vb8z
INDY Week
I Salvaged $6,000 of Luxury Items Discarded by Duke Students. Why Did It Make Me Feel So Terrible?
At the end of the school year, a lot gets tossed by Duke students at my downtown Durham apartment building. This year I decided to dive in.
Show HN: My LLM CLI tool can run tools now, from Python code or plugins (🔥 Score: 152+ in 3 hours)
Link: https://readhacker.news/s/6vbT2
Comments: https://readhacker.news/c/6vbT2
Link: https://readhacker.news/s/6vbT2
Comments: https://readhacker.news/c/6vbT2
Simon Willison’s Weblog
Large Language Models can run tools in your terminal with LLM 0.26
LLM 0.26 is out with the biggest new feature since I started the project: support for tools. You can now use the LLM CLI tool—and Python library—to grant LLMs from …
Why Cline doesn't index your codebase (Score: 150+ in 15 hours)
Link: https://readhacker.news/s/6vaJ2
Comments: https://readhacker.news/c/6vaJ2
Link: https://readhacker.news/s/6vaJ2
Comments: https://readhacker.news/c/6vaJ2
Cline
Why Cline Doesn't Index Your Codebase (And Why That's a Good Thing) - Cline Blog
Here's a common question we get from prospective Cline users: "How does Cline handle large codebases? Do you use RAG to index everything?"
It's a reasonable question. Retrieval Augmented Generation (RAG) has become the go-to solution for giving AI systems…
It's a reasonable question. Retrieval Augmented Generation (RAG) has become the go-to solution for giving AI systems…
Tariffs in American History (❄️ Score: 150+ in 3 days)
Link: https://readhacker.news/s/6v2d7
Comments: https://readhacker.news/c/6v2d7
Link: https://readhacker.news/s/6v2d7
Comments: https://readhacker.news/c/6v2d7
Imprimis
Tariffs in American History
When Alexander Hamilton became the nation’s first Secretary of the Treasury, he immediately began to prepare a schedule of tariffs, along with excise taxes on such commodities as alcohol and tobacco. The Constitution forbids taxing the exports of any state…
Show HN: AutoThink – Boosts local LLM performance with adaptive reasoning (🔥 Score: 153+ in 3 hours)
Link: https://readhacker.news/c/6vcs8
I built AutoThink, a technique that makes local LLMs reason more efficiently by adaptively allocating computational resources based on query complexity.
The core idea: instead of giving every query the same "thinking time," classify queries as HIGH or LOW complexity and allocate thinking tokens accordingly. Complex reasoning gets 70-90% of tokens, simple queries get 20-40%.
I also implemented steering vectors derived from Pivotal Token Search (originally from Microsoft's Phi-4 paper) that guide the model's reasoning patterns during generation. These vectors encourage behaviors like numerical accuracy, self-correction, and thorough exploration.
Results on DeepSeek-R1-Distill-Qwen-1.5B:
- GPQA-Diamond: 31.06% vs 21.72% baseline (+43% relative improvement)
- MMLU-Pro: 26.38% vs 25.58% baseline
- Uses fewer tokens than baseline approaches
Works with any local reasoning model - DeepSeek, Qwen, custom fine-tuned models. No API dependencies.
The technique builds on two things I developed: an adaptive classification framework that can learn new complexity categories without retraining, and an open source implementation of Pivotal Token Search.
Technical paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5253327
Code and examples: https://github.com/codelion/optillm/tree/main/optillm/autoth...
PTS implementation: https://github.com/codelion/pts
I'm curious about your thoughts on adaptive resource allocation for AI reasoning. Have you tried similar approaches with your local models?
Link: https://readhacker.news/c/6vcs8
I built AutoThink, a technique that makes local LLMs reason more efficiently by adaptively allocating computational resources based on query complexity.
The core idea: instead of giving every query the same "thinking time," classify queries as HIGH or LOW complexity and allocate thinking tokens accordingly. Complex reasoning gets 70-90% of tokens, simple queries get 20-40%.
I also implemented steering vectors derived from Pivotal Token Search (originally from Microsoft's Phi-4 paper) that guide the model's reasoning patterns during generation. These vectors encourage behaviors like numerical accuracy, self-correction, and thorough exploration.
Results on DeepSeek-R1-Distill-Qwen-1.5B:
- GPQA-Diamond: 31.06% vs 21.72% baseline (+43% relative improvement)
- MMLU-Pro: 26.38% vs 25.58% baseline
- Uses fewer tokens than baseline approaches
Works with any local reasoning model - DeepSeek, Qwen, custom fine-tuned models. No API dependencies.
The technique builds on two things I developed: an adaptive classification framework that can learn new complexity categories without retraining, and an open source implementation of Pivotal Token Search.
Technical paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5253327
Code and examples: https://github.com/codelion/optillm/tree/main/optillm/autoth...
PTS implementation: https://github.com/codelion/pts
I'm curious about your thoughts on adaptive resource allocation for AI reasoning. Have you tried similar approaches with your local models?
Why the original Macintosh had a screen resolution of 512×324 (Score: 150+ in 10 hours)
Link: https://readhacker.news/s/6vbLv
Comments: https://readhacker.news/c/6vbLv
Link: https://readhacker.news/s/6vbLv
Comments: https://readhacker.news/c/6vbLv
512 Pixels
Why the Original Macintosh Had a Screen Resolution of 512x342, and not 512x384 as One Might Expect - 512 Pixels
Many classic Macs came with — or supported — displays running at 512×384 pixels, but many compact Macs, ranging from the original 1984 machine up through 1991’s Macintosh Classic II had built-in CRTs running at 512×342 pixels. That covers all black-and-white…
'Strange metals' point to a whole new way to understand electricity (❄️ Score: 150+ in 2 days)
Link: https://readhacker.news/s/6v4Ee
Comments: https://readhacker.news/c/6v4Ee
Link: https://readhacker.news/s/6v4Ee
Comments: https://readhacker.news/c/6v4Ee
www.science.org
‘Strange metals’ point to a whole new way to understand electricity
Exotic materials with bizarre electron behavior could pave the road to revolutionary technologies
The Level Design Book (❄️ Score: 151+ in 2 days)
Link: https://readhacker.news/s/6v4mp
Comments: https://readhacker.news/c/6v4mp
Link: https://readhacker.news/s/6v4mp
Comments: https://readhacker.news/c/6v4mp
Leveldesignbook
The Level Design Book | The Level Design Book
The Level Design Book gathers level design knowledge for 3D video games in an approachable, up-to-date, and critical way. It is for designers of all experience levels and game engines.
The Difference Between Downloading and Streaming (Score: 150+ in 1 day)
Link: https://readhacker.news/s/6v8Ra
Comments: https://readhacker.news/c/6v8Ra
Link: https://readhacker.news/s/6v8Ra
Comments: https://readhacker.news/c/6v8Ra
Dan Q
The Difference Between Downloading and Streaming
Mostly to make use of an animation I made for a different blog post that hasn't materialised, let me take a moment to explain to you the fundamental technical difference between 'streaming' and 'downloading'. Which won't take long... because there isn't one!
The Ingredients of a Productive Monorepo (❄️ Score: 150+ in 2 days)
Link: https://readhacker.news/s/6v4kp
Comments: https://readhacker.news/c/6v4kp
Link: https://readhacker.news/s/6v4kp
Comments: https://readhacker.news/c/6v4kp
Negotiating PoE+ Power in the Pre‑Boot Environment (Score: 150+ in 11 hours)
Link: https://readhacker.news/s/6vcdj
Comments: https://readhacker.news/c/6vcdj
Link: https://readhacker.news/s/6vcdj
Comments: https://readhacker.news/c/6vcdj
The Polymath's Playbook
Negotiating PoE+ Power in the Pre‑Boot Environment
Solving a boot-time power challenge on PoE+ x86 systems by creating our own UEFI app to handle LLDP power negotiation at the firmware level, allowing our machines to request higher power from network switches before the OS loads.
The Hobby Computer Culture (❄️ Score: 151+ in 3 days)
Link: https://readhacker.news/s/6v2zM
Comments: https://readhacker.news/c/6v2zM
Link: https://readhacker.news/s/6v2zM
Comments: https://readhacker.news/c/6v2zM
Creatures of Thought
The Hobby Computer Culture
[This post is part of “A Bicycle for the Mind.” The complete series can be found here.] From 1975 through early 1977, the use of personal computers remained almost exclusively the province of …
Look Ma, No Bubbles: Designing a Low-Latency Megakernel for Llama-1B (Score: 151+ in 11 hours)
Link: https://readhacker.news/s/6vcet
Comments: https://readhacker.news/c/6vcet
Link: https://readhacker.news/s/6vcet
Comments: https://readhacker.news/c/6vcet
hazyresearch.stanford.edu
Look Ma, No Bubbles! Designing a Low-Latency Megakernel for Llama-1B