Knowledge catalogue
Mark Gurman / Bloomberg: Sources: Apple plans to launch Apple Intelligence in China in December via iOS 27.3, with features first rolled out in 2024; the launch won't include Siri AI — Apple Inc. is g
[UPDATE / RETRACTION: The '100% GTO' claim had a prompt-leakage bug] We patched the repo to pass strictly neutral states, here are the real, honest numbers: • Latency: Still ~2ms (ONNX + Absorbing Mar
After much testing and talking to actual users across all three proactive AI agents: Muse - best one for your parents, heavy IG/FB users, people who like or need easy setups, people who don’t want to
Banger paper from Google DeepMind on co-evolving an agent's harness and its weights. (bookmark it) If you already log your harness changes and results, those logs can become training data for the mode
My project is optimising qwen flash next on Mac Mini m5 64gb through SSD streaming. I recently asked LocalLLama for guidance on how to optimise token output. Today a popular method to do this is a loo
Brilliant paper on agentic continual learning from Meta and MIT. (bookmark it) Agent memory harnesses usually only write forward. A lesson goes into the bank and is never checked against whether it he
Character.AI, one of the first breakout startups of the generative AI boom, has filed for bankruptcy and is selling some of its assets to Disney. The company, which builds various online bots for user
I started Flama in 2018 to take a trained model to a production API with as little ceremony as possible. With 2.0 LLMs became first-class alongside classic models, and together with Miguel Durán-Olive
I built a @Grok @Bot that watches SPCX on its own — fully automated, no manual work. I’ve been running it for over a month and fine-tuned it several times. This version is very good. Every weekday bef
tl;dr: running Qwen3.8FN Q4 locally is cheaper for me than every cloud model except Luna and the newer Haiku at mid-peak and peak electricity pricing where I live, and becomes the absolute cheapest mo
Quartermaster (my local inference manager: one server that swaps llama.cpp, image, audio and video models in and out of VRAM) can now help you server and manage decision models: Auto-discovered. Drop
Hi, I see that vllm released d3 (a family of decision models) and llama.cpp already has support for them. I want to try them out. Are decision models sensitive to quantization? I know that for autocom
I’ve been building SonicBoom, a native tensor runtime written in C++/CUDA with a C ABI and no Python anywhere in the stack. I just tagged v0.0.1, and the first target is local LLM inference on consume
arXiv:2610.13303v1 Announce Type: cross Abstract: LLM judges are increasingly deployed as panels, and disagreement among panel members is often read as a symptom of an unclear rubric. We test a label-
arXiv:2610.13623v1 Announce Type: cross Abstract: Professional agents increasingly work through command-line interfaces and multiple tools over shared workspaces containing files, messages, and busine
We're launching Precision Numerics by releasing Boole-1.7B, a 1-bit version of Qwen3-1.7B. 247 MB packed weights: 13.9× smaller than BF16. 85.9% of the BF16 model’s average score across our 10-task ev
Ha gave this take to our team last week Will share some (very basic) stuff I vibe coded around this shortly Lots to be explored here! One of my more controversial bets: the AI feed will be the form fa
Must-read paper on the effectiveness of multi-agent systems and persistent memory. We have recently seen many viral examples of multi-agent systems scaled via persistent memory. This work takes a clos
New paper from Microsoft and colleagues on generating enterprise data for agents. If you train or evaluate agents on long-running business work, it's worth your time. (bookmark it) MiniCorp is an offi
Recommended NVIDIA paper on memory-efficient post-training RL. A very interesting technique for scaling RL for LLMs. The main takeaway is that you can update all of a model's weights with RL and still
arXiv:2610.13083v1 Announce Type: cross Abstract: Model providers now face distillation campaigns run through thousands of accounts, each of which looks like an ordinary user. When network and payment
My Jetson Orin–optimized engine, little-gemma V1.0, substantially outperforms llama.cpp. Even after exhausting every practical GGUF option, however, it still falls short of the ideal performance level
I run local LLMs pretty much around the clock, but I also enjoy gaming on the same PC. Having models sitting in VRAM while gaming isn't exactly ideal, and constantly unloading and reloading them manua
No more excuses for GPU-poor folks not to start LLMing! Full 27B-class reasoning in ternary transformer weights, for llama.cpp (CUDA, Metal, CPU) ~9.3x smaller than FP16 (ideal) | 98.2% of FP16 intell
Banger paper from Meta on AI research agents. (bookmark it) If you build AI-scientist systems, this one is worth your time. The result: A 27B open model beats the strongest open autoresearch baseline
Basalt: Blackwell inference engine Hi all! So over the last week I've been working on a Strata fork that's heavily tuned and can achieve throughput up to 2.6x what Strata usually does on the same weig
Spent the past month tweaking and experimenting with many different numbers to achieve 30tps. Hardware: -Rx6700xt 12gb vram (AMD) -2x16 ddr4 3200 ram -r5 5600x -llama.cpp vulkan sdk -window11 (no wsl
Pay close attention to custom harnesses. This is a super interesting paper showing the impact of training inside the harness. Qwen3-14B goes from 22.2% to 54.8% on Spider 2.0-SQLite when it is trained
Recommended paper from Sakana AI on recursive self-improvement. They propose an interesting way to scale recursive self-improvement through multi-agent self-supervision. In this line of research, self
Banger paper from Meta Superintelligence Labs on self-improving agents. (bookmark it) It's hard to know exactly what drives self-improvement, since so many variables are at play (notes, skills, tool c
Build your own harness, folks. Reading papers like this makes me realize how underexplored harness engineering really is. The authors find that on whole-repository migration, GPT-5.6 Sol goes from 6.5
Disclosure: I work at H2O.ai. We released H2O-Lightning-4B, an open-weight (Apache-2.0) model for the 'decisions API' style of inference that Jev made popular: you send a state plus typed questions (p
Hi everyone! So far in 2026 I have been iterating and evolving a research system I designed to help me sift through things that LLMs all agree on that may not be correct, and to give me actual confide
LM Studio 0.4.26 is out now! • ⚡️ `/v1/systemone` and `/v1/decisions` APIs • 👾 Coexistence with Bionic • 🎨 UI updates and bug fixes Yes, you can now point your Bionic agent to models loaded in your LM
All anyone ever talks in this sub is about x new model having x benchmark scores and how well it can code. Or some new tech to increase t/s for qwen. So I wanted to go back to the roots of this sub an
Preamble and disclaimer, Sample size: 8, at any size of form this research is just screwing around being writing it down. This was just a little,(well... big), curiosity test I wanted to run to see wh
Another great paper from Salesforce AI Research. The finding is that a 31B open Gemma-4 web agent scores 74.6% on the 9-app WebArena Infinity set, above Gemini 3 Flash with browser use at 70.1%. They
Anthropic is making changes to its usage policy for the first time in over a year to reflect new and high-risk cases of misuse - including election interference, weapons development, surveillance, and
Google is launching a 'universal' Gemini AI agent that can work across apps and devices in the background. The tool, announced as part of the Gemini at Work event on Thursday, will be available within
Great paper from Google on program repair agents inside CI. (bookmark it) They propose FlowAgent, which runs a ReAct-style generate-and-validate loop on pre-submit test failures and shows the fix insi
arXiv:2610.09000v1 Announce Type: cross Abstract: As small language models (SLMs) are increasingly deployed on resource-constrained and on-device platforms, including as components of agentic systems,
In recent decades, Ireland has grown into a vibrant hub for global technology. As modernization accelerates, Irish organizations — spanning government agencies, established enterprise brands, and grow
The other week I posted about Jev vs. Kev compared and since then, OpenAI released the decisions endpoint, Cloudflare released Clef and many here asked about Laya as well. This time we compared six po
arXiv:2610.09981v1 Announce Type: cross Abstract: LLM routers pick a cheap or expensive model per request by its content, and many gateways and some cloud platforms can log that choice with content lo
Hi all, having searched google and asked several AIs without success, maybe s.o. can help. I have 2 x dgx spark in a cluster. deepseek-flash is running successful. Now I want to test qwen3.8-flash-nex
arXiv:2610.10411v1 Announce Type: cross Abstract: Speculative decoding accelerates large language model inference by using a low-cost draft model to propose tokens that the full-size target model veri
Banger paper from NVIDIA. One exciting trend I am seeing is building verifiable environments for your agents and training the harness alongside the model. Co-evolving the harness and the model is a bi
On 15 September, TypeSafe released Jev, and within a few weeks about 10 companies had shipped something that works the same way. OpenAI, Liquid, Cloudflare, AWS and PostHog all launched one, and an op
arXiv:2610.07094v1 Announce Type: cross Abstract: LLM deployment is shifting from single-turn completion to agentic trajectories in which a model plans, calls tools, reads results and reasons at test
arXiv:2610.07447v1 Announce Type: new Abstract: Autoresearch agents tackle open-ended problems by repeatedly proposing candidate solutions, evaluating them, and using feedback to guide subsequent expe
arXiv:2610.08414v1 Announce Type: new Abstract: Image Bitstream Fine-grained Understanding (IBFU) aims to directly perform fine-grained classification and semantic description generation from encoded
arXiv:2610.07519v1 Announce Type: new Abstract: Automatic sign language translation (SLT) has entered consumer products, turning American Sign Language into English text for dictation, messaging, and
arXiv:2610.07026v1 Announce Type: new Abstract: The Offline AI Modules workstream enables practical, low-power, and community-accessible deployment of voice-first AI systems that operate fully offline
arXiv:2610.07935v1 Announce Type: new Abstract: LLM agents are increasingly capable of executing complex tasks and of recursively improving themselves on easy-to-verify objectives such as software eng
Co-authors:Stenal Jolly, Strategic Cloud Engineer, GoogleAnubhav Dhawan, Software Engineer, Google Following the landmark announcement of MCP Toolbox v1.0, we're thrilled to announce that the MCP Tool
Banger paper from Meta AI on research agents that decide what to investigate next. (bookmark it) If you run long-horizon research agents, choosing the next investigation is hard to learn, because thos
Disclosure: This unit was provided free of charge by Orange Pi for testing. No editorial review, no preconditions, no script. All data, bottlenecks, and thermal behavior are reported directly from har
Off-the-shelf AI assistants forget you between conversations. This post shows how to build a personal assistant that accumulates context using OpenClaw on Amazon Bedrock AgentCore runtime, with AgentC
Standard distillation usually means burning weeks of compute and billions of tokens hoping the student model eventually mimics the teacher. We wanted to see what happens if you skip backprop entirely
Context compression is a huge bottleneck for long-running agents. This work finds that context compression hurts long-horizon agents at a few specific points. They propose PAIR, which replays the agen