High-quality GLM-5.2 Quant on 4x DGX Spark – Guide, Results, and Comps
I got GLM-5.2 NVFP4 running on four DGX Sparks at 128K context. This is still a niche/hacky setup, but it is now a real serving point…
Latest
New meta
New metaCitrini: It might be time to consider a long “likely to get approval for latest model” companies vs short index trade
Third-parties should focus on scrutinising system cards
By default, I expect system cards will get worse, which would be bad. Some mechanisms could improve system cards, but I expect they will be outweighed.…
> Completely trained from scratch new models will be released by @SpaceX every month this year. Very aggressive, but happy to see someone committing t…
> Completely trained from scratch new models will be released by @SpaceX every month this year.Very aggressive, but happy to see someone committing to itElon Musk:…
MLX Fine-Tune Example Guide
A Local MLX Fine-Tuning Experiment Just finished a local LoRA fine-tune of a 7B instruction model on Apple Silicon, via MLX, teaching it a high-fantasy literary…
b9837
jinja, chat: add --reasoning-preserve flag (#25105) jinja, chat: add --reasoning-preserve flag correct help message macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED…
RT Palmer Luckey: At least 1,200 people died of heatstroke in France last Wednesday, mostly elderly people indoors. Even crazier, about 70,000 Europea…
RT Palmer LuckeyAt least 1,200 people died of heatstroke in France last Wednesday, mostly elderly people indoors.Even crazier, about 70,000 Europeans die from the same every…
I built an agent Harness for Small Models. I got Qwen 3.5 4b managing servers.
This is something I've been working on, I like playing around with smaller local models but found most agent harness's not well suited for them. The…
Faster Gemma 4 on MLX with multi-token prediction
Gemma 4 is now significantly faster in Ollama 0.31 on Apple Silicon via multi-token prediction (MTP), powered by MLX. Performance is now up to 90% faster…
Introducing the Claude apps gateway for Amazon Bedrock and Google Cloud
Introducing the Claude apps gateway for Amazon Bedrock and Google Cloud
Claude in Microsoft Foundry is now generally available
Claude in Microsoft Foundry is now generally available
RT Sakana AI: TechCrunch on Sakana Fugu and the broader conversation around frontier AI access in Asia. Our position remains that AI is best developed…
RT Sakana AITechCrunch on Sakana Fugu and the broader conversation around frontier AI access in Asia.Our position remains that AI is best developed together rather than…
TechCrunch on Sakana Fugu and the broader conversation around frontier AI access in Asia. Our position remains that AI is best developed together rath…
TechCrunch on Sakana Fugu and the broader conversation around frontier AI access in Asia.Our position remains that AI is best developed together rather than hoarded. Sakana…
True
TrueHouse Republicans: 100 million dead.That’s the verdict on Marxism. History already ran this experiment.Never again.
Locally running mode turns an Image into a Cute Controllable Character you can Play as
This is a sequel to my last post here !! It meant a lot to have such positive feedback last time. This is the 800M version…
Better Images of AI
Article URL: https://betterimagesofai.org/ Comments URL: https://news.ycombinator.com/item?id=48713051 Points: 5 # Comments: 1
We need tech news sources which exclude AI
Its now clear that we need to preserve tech press for non AI related things. Techmeme for example is now completely overrun with AI stories.HN is…
RT Grummz: Happy Birthday to Elon Musk. Thank you for buying Twitter. It saved free speech.
RT GrummzHappy Birthday to Elon Musk.Thank you for buying Twitter. It saved free speech.
AI Daily Brief — 28 June 2026
Coinbase halved its AI spend by routing to Chinese models including GLM 5.2, as Semgrep's benchmark shows GLM 5.2 outperforming Claude in cybersecurity.
RT Yohei: embedded the AI Engineer World Fair talks to be able to “find similar talks”, but then added a map & ability to bookmark a list and export…
RT Yoheiembedded the AI Engineer World Fair talks to be able to “find similar talks”, but then added a map & ability to bookmark a list…
Powerful qualities in type B physicalism
Type B physicalism is a view in philosophy of mind, and "powerful qualities" (PQ) theory is a view in the metaphysics of properties. They share an…
Qwen3.6-27B UD Q3 with kv at q8 is quite amazing for simple proof of concepts
Preface, technology is not my industry, but I am a very passionate poor man. So much so that I discovered 'AI' - ChatGPT in the beginning…
NPC Engine Using Local Models
I’ve been working on a game-agnostic NPC engine/backend based pretty heavily on SillyTavern-style architecture, and with smaller local models getting better and better, I honestly think…
On cigarettes
Article URL: https://funnelfiasco.com/blog/2026/06/28/on-cigarettes/ Comments URL: https://news.ycombinator.com/item?id=48712762 Points: 32 # Comments: 37