Fixing Tool-Call Accuracy in vLLM(ubicloud.com)
20 points by coffeemug | 1 comments
Ubicloud blog post explaining tool-call schema accuracy problems when serving open-weight models like Kimi K2.6 in vLLM and how constrained decoding fixes them.
Hacker News front page, screenshotted
Updated , every 15 minutes
ai-agentsllmopen-sourcerustanthropicclaudeopenaiphppythonprivacyclaude-codegithubcvemetareverse-engineeringai-coding-agentscppmcpbenchmarkscertificate-authoritygooglemathematicsandroidapple
Switzerland. Block by block. At 1:1 scale(blocks.manifold3d.ch)
299 points by theanonymousone | 133 comments
Product page for a 1:1 scale Minecraft Java map of Switzerland built from swisstopo open geodata, with a free Jungfrau-Aletsch area and paid full-country download.
Birdwatching rewires the brain and slows cognitive decline(bbc.com)
202 points by cisc | 78 comments
BBC Future article on research showing birdwatching reshapes the brain, may slow cognitive ageing, and is used as therapy for dementia patients.
Tasks as Code: Move Your Tickets into the Repository(typemill.net)
32 points by trendschau | 25 comments
A Typemill knowledge-hub article arguing for storing development tasks as Markdown files in Git so AI coding agents can read them alongside code and docs.
One of the Greatest Inventions in the History of Human Transport(humantransit.org)
37 points by skadamat | 8 comments
Human Transit blog post by Jarrett Walker on Blaise Pascal's 1662 Paris carriage network, the first fixed-route public transport, and its network design principles.
Show HN: TitleArt – Make text like it's 1997(titleart.dvy.io)
302 points by davidbarker | 54 comments
TitleArt is a browser-based WordArt-style editor with a Windows 9x interface for creating text with classic shapes, colors, shadows, and 3-D effects.
Anthropic S-1(sec.gov)
326 points by SiNTEx | 259 comments
Anthropic, PBC's Form S-1 registration statement filed with the SEC for an initial public offering of Class A common stock on Nasdaq under the symbol ANT.
Behind the scenes at the NeXT Computer office(exhibits.stanford.edu)
173 points by Austin_Conlon | 96 comments
Stanford Libraries exhibit gallery of Doug Menuez photographs documenting daily life inside NeXT Computer, Inc.
Websites Should Be Printable(brennenputh.me)
100 points by bagatelle | 92 comments
A personal blog post arguing websites should be printable, sharing the author's @media print CSS snippet for hiding headers, adjusting font sizes, and styling links.
PaperPilot – Zero-Cloud, Pure-Rust PDF Engine with WASM and MCP(github.com)
187 points by vardhan1237 | 69 comments
GitHub repository for PaperPilot, a local, zero-cloud pure-Rust PDF engine and AI document copilot with WASM builds and an MCP server.
Dolly Parton's Imagination Library(imaginationlibrary.com)
592 points by jumploops | 99 comments
Homepage of Dolly Parton's Imagination Library, a nonprofit that mails free books monthly to children from birth to age five in five countries.
Growing Trees(brashandplucky.com)
128 points by ibobev | 18 comments
Blog post explaining an implementation of procedural tree growth based on Palubicki et al.'s self-organizing tree models, covering light competition, resource allocation, and foliage rendering.
Overgrown, when plants invade your abandoned tabs(charliegerard.dev)
347 points by speckx | 118 comments
Blog post by Charlie Gerard introducing Overgrown, a Chrome extension that grows plants over neglected browser tabs which users must slash away to access them.
Colorfigures(april-zhu.com)
9 points by the-mitr | 2 comments
Landing page for Colorfigures, an MIT-licensed AI agent skill that generates interactive, hand-inked isometric HTML figures of everyday machines from a text prompt.
Show HN: A Truly Modular Keyboard(github.com)
211 points by NotARoomba | 69 comments
GitHub repository for Keybytes, an open-hardware modular keyboard where each key is its own PCB with a RISC-V MCU, connected via pogo pins and magnetic cases.
Attention Is Not What You Need (2025)(arxiv.org)
148 points by westurner | 53 comments
arXiv paper proposing an attention-free sequence model using Grassmann flows and Plucker coordinates, compared against Transformers on Wikitext-2 and SNLI.
Writing Is Closer to Thought Than Code(danielmiessler.com)
23 points by eigenBasis | 7 comments
Daniel Miessler blog post arguing that written intent, via an 'Ideal State Artifact', rather than code should be the source of truth when building software with AI.
Iron Age salt-mining complex found beneath Hallstatt in Austria(archaeologymag.com)
98 points by janandonly | 42 comments
Archaeology News article reporting the discovery of a large Iron Age salt-mining chamber complex, at least 300 meters long, beneath Hallstatt, Austria.
Why I Am a Programmer(alfie.wtf)
275 points by jruohonen | 85 comments
Personal blog essay recounting how BBS-era cracktros, the demoscene, and Denthor's ASPHYXIA Pascal demo tutorials inspired the author to become a programmer.
Superhuman Hardware Design Is the Easy Part(a16z.news)
46 points by gmays | 31 comments
An a16z essay arguing that AI models may soon master hardware design, but the lossy interface between digital designs and physical manufacturing and measurement remains the hard part.
We should be more ambitious with the software we build(danluu.com)
840 points by ltratt | 726 comments
Dan Luu essay arguing that advancing AI coding models let developers build more ambitious software, citing pgrust, a regex engine, ripgrep fork, and board game AI projects.
Steal My Bitcoin CTF PoC(github.com)
141 points by davikr | 38 comments
Anthropic S-1(sec.gov)
14 points by ChrisbyMe | 0 comments
The State of Context Engineering(dbreunig.com)
218 points by pchristensen | 56 comments
Slides and transcript of Drew Breunig's AGNTCon talk "Context Fails & Fixes," reviewing the history, failure modes, and future of context engineering for LLMs.
IBM CEO says open weight models are now good and cheap on-prem(theregister.com)
352 points by Betelbuddy | 252 comments
The Register reports IBM CEO Arvind Krishna's Q3 earnings call remarks that open weight models now offer 90% of frontier performance at a fraction of the cost for on-prem use.
Amazon’s “digital pacifier” for toddlers(oligarchwatch.substack.com)
240 points by jumploops | 344 comments
Oligarch Watch article criticizing Amazon's new Kids Tablets marketed to toddlers, citing pediatric research linking screen use to developmental and emotional problems.
F-Zero X on Nintendo 3DS(hypnotize.works)
143 points by bpierre | 80 comments
A post-mortem of porting the F-Zero X decompilation (G-Diffuser) to Old and New Nintendo 3DS with a native renderer and 64DD Expansion Kit, built largely by Claude agents.
Two Million Hornets(dark-mountain.net)
30 points by promulgate | 4 comments
An essay by Tom Smith from Dark Mountain Issue 30 about trapping invasive Asian hornets, wildfires, and changing mountain communities in northern Spain.
A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis(lynchlab.net)
613 points by sungam | 259 comments
A dermatologist's blog post comparing two frontier LLMs with a fine-tuned Inception v3 CNN on diagnosing 200 biopsy-proven skin lesions, where the CNN scored 62% versus the LLMs' 42%.
Show HN: ADRs for AI – Deliberation records for human–AI work, versioned in Git(adrs.systems)
59 points by lumbroso | 32 comments
Landing page for ADRs4AI, a Git-versioned framework, template, VS Code extension, and iOS app for recording human–AI design deliberations as decision records.