Knowledge catalogue
I have enabled the vision for the CIRU Strix UL4 quant of Qwen 3.8 flash next (others quants likely perform very similar) and tried it on a few things, then wanted to show my partner how great it work
Interesting paper to improve recurrent reasoning. Looped models are great because you get more reasoning out of a model without adding parameters. So this work proposes a looped architecture with a ne
I've seen a lot of Super Mario showcases, but never a mashup or derivative work. So I took a Super Mario generation prompt from ChopSticksPlease's Post here - and added a Matrix twist - and found myse
Hola all. Do you guys mind sharing your LLama.cpp config and system setup details for Qwen3.8 Flash Next? Model's quite big and tryining many combinations of llama.cpp options takes lots of time, so l
last night I was messing around with Codex and threw in a stupid prompt: **“hey codex, find me a girlfriend. make no mistake”** I expected the usual AI answer. You know, go outside, join a hobby club,
Very interesting paper if you are building with agents. How should a website tell an AI agent what it may access, for what purpose and at what price? robots.txt can only allow or disallow paths. It ca
What happens if you agents to run a town's economy? This super interesting paper provides some insights: They put 100 LLM agents in charge of a town economy for 26 simulated weeks, and they find that
I have been battling with my Qwen3.8:27b setup on my rtx 5080 16gb. I am using llama.cpp to run a nvfp4 version of qwen3.8:27b llama-b10699-bin-win-cuda-13.3-x64llama-server.exe -hf williamliao/Qwen3.
arXiv:2609.10611v1 Announce Type: cross Abstract: Membership inference attacks (MIAs) have emerged as critical tools for auditing privacy risks in large language models (LLMs), aiming to determine whe
arXiv:2609.09647v1 Announce Type: new Abstract: Agentic systems are rapidly moving to production, where they read untrusted inputs, call tools with real permissions, and act autonomously, expanding th
Great paper on self-evolving agent harnesses. Self-evolving agent harnesses have two practical problems: 1. Search is slow, because every candidate harness needs repeated agent runs and code edits. 2.
Harness engineering is a top skill right now This new Meta paper is a good production example. Auto-RecSys runs autonomous research on Meta's industry-scale recommendation models, where one training r
Interesting paper from Microsoft. If you run persistent memory for a production agent, this one is worth your time. (bookmark it) Memory curators usually read only the finished trajectory. That lets t
arXiv:2609.10385v1 Announce Type: cross Abstract: Animation and VFX pre-production review requires teams to translate loosely specified creative intent--briefs, evolving specifications, heterogeneous
arXiv:2609.11490v1 Announce Type: cross Abstract: An unlearning audit reads its verdict off numbers that an unlearned model and its retrained reference each publish, and both also ship batch-normaliza
Very cool paper on memory compression for agents. If you run many agent sandboxes in parallel for RL or evals, memory becomes highly redundant. This work suggests that compressing against that redunda
The downside of uncensoring a model is that it is known to potentially damage it, but CyberTiel is an even more capable software engineer than its censored TielCoder base, while allowing offensive sec
Great paper from AWS. I use a similar setup where an agent orchestrator sits on top of a multi-agent system. (bookmark it) This work introduces one of the many approaches available to manage compound
If you run automated prompt or harness evolution, this one is worth your time. (bookmark it) Semantic-gradient prompt optimization improves an agent harness from execution feedback, but its local upda
Spent about two weeks convinced I had a prompting problem. My local Qwen3 agent would just stop. No error, no crash, finish_reason came back as 'stop' like the model had decided it was done. Except it
Model available here: Qwen3.8-27B-Uncensored-Genesis-V1-MTP-GGUF This model is a practical realisation of things described in this paper, but adapted by me for machine learning: https://arxiv.org/pdf/
arXiv:2609.06893v1 Announce Type: cross Abstract: Direct preference optimization DPO is a promising offline approach for aligning large language models (LLMs) due to its simplicity, computational effi
Women came together in London to talk about building careers in AI, how they got here, what comes next, and what it takes to move forward in a field that’s changing fast. Replit and our partners at th
What is Astra actually supposed to be better at, Because if the answer is “everything,” then I genuinely do not understand what kind of testing was done for writing-heavy use cases. My biggest issue i
I'm the CTO of a mid-size nonprofit, and I run a local Qwen stack as a coding subagent for that work — partly on principle (a fair amount of what I handle shouldn't leave the building), partly because
I kept picking local models badly. Parameter count, quantization, VRAM. I could read all of it and still not know whether a model would be any good on my machine. So I stopped guessing and wrote somet
Privacy have been concern of many of us to have their own hardware to run llms, and here's another reason why: two mathematicians spent a year cracking one of the hardest problems in math and fed ever
Hi all! I just wanna say that I am tired lol. Yes, it's another harness, but I spent a lot of time and effort and have forsaken my hobbies to build the Jenny (like XJ-9) app. Jenny is a free, MIT lice
Banger paper from Google DeepMind. Every AI weather model so far has been trained and initialized on analysis data, which is itself the output of another model. That means the forecast inherits whatev
Banger paper from Microsoft and Cornell. If you have looked at thinking tokens and decided you cannot afford the context, read this one. (bookmark it) A pause token buys the model extra compute for ea
OK, you guys want to start with local vibecoding but don't know how. In the discussions, you recommended that I use Godot. I downloaded it without knowing what it was. I was expecting a library, but i
Hi, I am thinking about putting some money / time into my local ai learning path and 'cleared' my attic where I found the following hardware. 3 x NUC11 (Core i5 1145G7 / 2,6 GHz, 64 GB RAM DDR4 2666MH
arXiv:2609.03423v1 Announce Type: new Abstract: Full-duplex voice agents must continuously decide when to listen, backchannel, interrupt, handle speech overlaps, take the floor, and yield. Existing be
Great paper on designing multi-agent systems. How many distinct communication topologies does an LLM multi-agent system actually need? This works claims that it's about six. Technical summary: Researc
Great tips on working with reasoning models. Normally you would append what the model figured out after the document and ask again. It turns out that where you put the reasoning trace changes long-con
I am getting into developing a RAG setup, for getting information out of existing documents, new document ingestion, web searches, and good visuals. I am planning to use it for, alongside the regular
Insightful paper from Microsoft and colleagues. If you have ever had an agent run fail 80 steps ago with no way to find where, this one is for you. (bookmark it) Agent failures show up as long complex
arXiv:2609.03966v1 Announce Type: new Abstract: Agent evaluations report a tool-call rate read off the serving stack. That number can be zero while the model is emitting well-formed calls: the interfa
This work introduces diffusion-augmented LLMs, a new class of models. They first suggest that speculative decoding needs a separate draft model. Diffusion LLMs give up the quality of the model they re
Banger paper from BAAI. If you are building research agents, this one is worth your time. (bookmark it) They find that adding skills scores 134.3% higher on MLE-bench, 34.4% higher on PaperBench, 9.2%
Banger paper from Google DeepMind and colleagues. (bookmark it) A model reads its entire KV cache on every generated token, even though it ends up attending to a tiny slice of it. In other words, if y
Brilliant paper on long-horizon agents. They cut 78.9% of an agent's LLM calls while raising its success rate. Here is how: It turns out that ReAct issues one primitive action per model round. That al
Good measurement work on whether retrieved agent skills actually help. They report that agent skills that lift your aggregate score can be hurting every task they touch. The usual way of checking comp
Integrate Microsoft Outlook with Amazon Quick to automate email management, calendar scheduling, and workflow coordination. This post walks through the end-to-end setup and shows automation scenarios
It started as a fork of llama-swap, but I have been building it out for myself since then as a convenient tool for all my local AI needs, and by now it has drifted far enough to be its own thing. The
No. I still don't think that pure scaling of LLMs can lead to AGI. BUT... ... most recent advancements aren't actually about scaling LLMs. The really about adding deterministic, symbolic stuff to LLMs
Disclaimer: I'm the builder. --- Over the past 2 yrs, I was working on a SaaS ML project & got very interested in ML/DL/AI. As everybody else, there were some normal paths I took to build a solid unde
// Agent Zero Memory // This work separates three things that agent memory systems usually collapse into one. If you build agents with long-term memory, this memory design is a worth a read. Here is h
Banger paper from ByteDance Seed. If you are curious about self-evolving agent harnesses, this one is worth your time. (bookmark it) The proposes method, HarnessDev, stops scoring a model on the tasks
arXiv:2609.00365v1 Announce Type: new Abstract: Command-line coding agents (e.g., Claude Code, Gemini CLI) can already read and write files and sustain long sessions, yet end-to-end research still fra
Finally, a good paper testing if graph memory actually beats flat retrieval for long-term agents. (bookmark this one) Researchers extract each conversational turn into typed nodes and attributed edges
arXiv:2609.01046v1 Announce Type: cross Abstract: Production LLMs must handle inputs that attempt to override system instructions, bypass safety policies or elicit harmful responses. A common mitigati
Independent verifiers improve agent output - but frontier verifiers are expensive. So we trained our own. Results on @GoogleDeepMind's FACTS-Search: > No verifier: 89.4 (previously published SOTA) > W
arXiv:2609.00791v1 Announce Type: cross Abstract: In this work, we introduce Instella-MoE, a fully open Mixture-of-Experts (MoE) language model with 16 billion total parameters and 2.8 billion active
arXiv:2609.01272v1 Announce Type: new Abstract: Prospective memory means carrying out a deferred intention at the right future cue while other work continues. Benchmarks now isolate it as an agent ski
Nice paper with great insights on improving self-evolving agents. Self-evolving agents fail in three specific ways: 1. Terminal-only feedback makes it ambiguous which step caused the error. 2. Agents
arXiv:2609.00546v1 Announce Type: cross Abstract: Agent systems are commonly described by the model and harness that currently produce their behavior. That boundary is useful for one execution but und
Banger paper from the Qwen team. If you evaluate agents on anything longer than a single session, this one is worth your time. (bookmark it) E-Commerce Bench runs an agent through a simulated 365-day
arXiv:2608.29610v1 Announce Type: new Abstract: The current alignment tuning paradigm for Large Language Models (LLMs) prioritizes surface-level behaviors -- fluency, safety, and tonal consistency. Wh
Brilliant paper on reducing reward hacking in agents. If you follow the recent OpenAI <> HuggingFace incident, you might want to check this paper out. (bookmark it) The usual response to reward hackin