yaoke259
Re:Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5
on Wednesday October 28, 2026 @10:20PM (#50283806)
It could be interesting to experiment with more complex representations in the memory/file system, i.e. vectors. Seems like the full advantage of this architecture is yet to be explored. Blending neurons with deterministic compute to fit on one GPU is a fascination concept, never considered it before!
Re:Early progress on accelerating drug development with Claude
on Tuesday October 27, 2026 @07:47PM (#50267503)
Given that a lot of the bottleneck in speed is human trial, I wonder if overall speedup will be limited anyway?
Automating or streamlining human/animal trials with 24/7 monitoring and AI check-ins would be huge.
Re:Amazon Is Gutting Its AI Division After Sustained Failure
on Tuesday October 27, 2026 @07:34PM (#50267399)
> Instead, labor and computing power are being diverted toward a singular “frontier-model effort”
They are just changing their strategy to invest in a single model? It's not like their AI department is in shambles...
Re:Introducing LFM3: Every Modality for Specialized Intelligence
on Tuesday October 27, 2026 @07:22PM (#50267291)
First class hardware support makes this really exciting and different. On device processing is cheaper and faster, critical for real time applications of AI. Really excited for the rise of local, smart intelligence!
Re:Using LLM for confidence scoring is a bad idea
on Wednesday October 21, 2026 @11:36PM (#50190897)
Whilst general self-reporting of confidence will be hard to achieve (we can't even do it as humans), I'm sceptical of the notion that we can't train models to give some confidence scores accurately. I think over the past year that with the correct RL, models can showcase some pretty insane capabilities. If reporting the correct confidence was a RL training target say for predicting weather events or a game outcome that is supported by real RL environment or simulated data, then I suspect LLMs will develop a strong intuition for confidence similar to how the recent JEV and decision models are trained.
Re:The recently proposed o(n log n) integer multiplication algorithm [...
on Wednesday October 21, 2026 @11:30PM (#50190856)
Awesome write up and intro by a fellow Sydney-sider, kept me hooked! As someone who's not a maths wizard, the technical details flew by my head but the prose allowing me to empathise deeply with those in the field was very enlightening. I never realised how big of a deal the maths dump was and how much it has shaken up the field. I guess we are all just trying to figure out this whole "intelligence on tap" thing together.
Re:Launch HN: Alter (YC S25) – Access control for multi-user agents
on Wednesday October 21, 2026 @06:36PM (#50188659)
what's the advantage of a tool like this compared to existing mature tools like infisical?
Re:What shall I teach my son?
on Wednesday October 21, 2026 @02:38AM (#50176876)
> we can't tell the future for the next year- how could we possibly predict what works well 20 years from now?
well said, absolutely agree
Re:What shall I teach my son?
on Tuesday October 20, 2026 @11:54PM (#50175905)
I often think of this question. Probably things that are enduring. How to be a good person, how do people work, how does society work, how to learn, some hobbies and life skills.
I've been building voice agents for the past year, and honestly the word error rate and how good they sound are "good enough" across countless providers. The real gaps are in latency, handling interruptions, interrupting the speaker if needed (this is surprisingly useful in most contexts) and adapting to the emotion/tone of the context. Would love to see a benchmark or tests that measure these effectively and test the claims of Voxtral 3! These existing benchmark results look very impressive.
Re:DeepSeek just open-sourced its inference engine
on Tuesday October 20, 2026 @09:33PM (#50174989)
XD
Re:DeepSeek just open-sourced its inference engine
on Tuesday October 20, 2026 @08:40AM (#50165464)
in terms of raw token throughput and serving users it is, but in terms of R&D and long term lab capabilities, training is still the goose that lays the golden egg
Re:DeepSeek just open-sourced its inference engine
on Tuesday October 20, 2026 @07:13AM (#50164502)
Supporting Huawei Ascend 950 NPUs with open source inference code has huge implications in the global AI hardware market. Imagine if all smaller labs now can build on top of this stack and add increasing support for more compatible hardware , software and models. This could seriously challenge NVidias death grip. Although inference is still only a part of all AI workloads and open sourcing training on Huawei chips would be even bigger.
Re:The US orders every civilian agency to rip out DeepSeek, even tools...
on Tuesday October 20, 2026 @12:08AM (#50161246)
People love convenience. Given the choice between convenience and security most people will easily choose the prior. Security is also just hard. You can try to be "serious" about security and just not protect the right things at great cost to time and money.
Re:Ask HN: How to deal with shipping AI-written code you don't underst...
on Monday October 19, 2026 @09:10PM (#50160121)
For me personally, iterating is huge. You can't arrive at a hugely complex system in one shot. I think of the LLM as someone I'm running software engineering "experiments" with. As requirements become clearer, I discuss with it more and more and over time I'll gain a stronger understanding of the problem and possible solutions.
You'll also discover over time which systems the AI can handle fine on its own and which abstractions and systems it just has no ideas on how to design. Often I find I have no idea how to design those too. It's then useful to try different designs, test tradeoffs and refactor aggressively if needed.
Re:Token-maxxing is dead, OpenRouter's token usage plateaus
on Monday October 19, 2026 @09:06PM (#50160081)
Agreed, OpenRouter is competing in an increasingly crowded market and even a lot of cloud providers like Neon, Vercel and Cloudflare are rolling out their own gateways to vertically integrate with their hosting. It's hard to say what the true net usage of LLM look like in the world!
Re:Timestral: Mistral release time-series foundation models
on Monday October 19, 2026 @09:03PM (#50160063)
ChatGPT became massive because of it's generalisability without needing training a new model for every domain and for its zero-shot learning inference/learning capabilities. If this model can be scaled or easily fine-tuned, it can have huge implications on forecasting in domains where custom models are too hard or too expensive! I am somewhat sceptical and wonder if the big quant firms already have versions of this internally or if not, why no one has achieved these kinds of results until now.
Re:MiniMax M3.1 Flash
on Monday October 19, 2026 @08:55PM (#50160004)
what a bizarre benchmark, cool!
Re:Ask HN: What are your worst war stories bringing agentic applicatio...
on Sunday May 31, 2026 @06:52PM (#48350541)
what have you guys tried?
Ask HN: What are your worst war stories bringing agentic applications into prod
For a bit of context, I’m currently creating a team of AI agents at work to generate reports by fanning out into a large amount of subagents to process a large amount of transcript data. When the analysis fails mid-way because of some individual step like an API call returns an error or the machine is out of memory, it would create cascading errors that break the entire generation with almost...
Show HN: Using Claude Code history to predict your next messages while you sleep
Overnight is a free, open source CLI supervisor/manager layer that can run Claude Code (Codex support coming soon!) by reading your Claude conversation histories and predicts what you would’ve done next so it can keep executing while you sleep.Github (MIT license): https://github.com/yail259/overnight Landing Page: https://workovernight.com/What makes it different to all the other generic “run...
Show HN: Ptero, a Svelte Alternative to Docusaurus
https://github.com/yail259/ptero
Hey guys, happy to share a passion project I’ve been working on. It’s called Ptero (short for pterodactyl, get it? haha) and is a Svelte based alternative to Docusaurus so you can now build your dev docs page in svelte! * Github: https://github.com/yail259/ptero * Landing page & dogfood demo: https://ptero.yaoke.pro/ Is this project necessary when Astro, Docusaurus and all the other...
Show HN: TypeScript runtime that syncs state for your multiplayer app/game
I’ve been experimenting with Phaser for small web games and found that even simple multiplayer logic becomes complicated fast. I built Martini-Kit (https://github.com/BlueprintLabIO/martini) to reduce that overhead.It’s open source, written in TypeScript, and lets you structure your game logic almost like a single-player project while the library handles state syncing. It provides first-class...
Re:Show HN: MTXT – Music Text Format
on Thursday December 04, 2025 @07:32AM (#46146931)
pretty cool!
Show HN: Martini-Kit, create multiplayer games without writing networking code
https://martini.blueprintlab.io/
I was working on a multiplayer racing game and kept hitting the same issues. State desyncs where players would see different positions. Race conditions when two players interacted with the same object. The usual stuff.The frustrating part was that these bugs only showed up with multiple real players. Can't reproduce them locally, can't easily test fixes, and adding logging changes the timing...
Re:Liquid Glass UI
on Monday October 13, 2025 @12:56AM (#45564802)
would be cool but rn just looks like normal glassmorphism?
Re:AI getting gold in IMO is not that surprising
on Monday October 13, 2025 @12:55AM (#45564795)
First blog post on HN!
AI getting gold in IMO is not that surprising
Re:Show HN: Prompts to make AI impersonate 10 programming legends
on Saturday October 11, 2025 @05:20AM (#45547732)
Why? It's like saying celebrity impersonations is unethical. Why is impersonating someone inherently bad?