Interesting Finds — 2026-09-16
Five notes: Tailcat's accountless tunnels, Claude Money's trust question, System One models that cannot hallucinate, recursive self-improvement through dreaming, and an inbox for long-running agents.
tag — research
Interesting Finds — 2026-09-10
Five notes: Nvidia's Hydra-0 action-flow world model, a random human life, Google's Putty vibe-coding experiment, Qwen-RobotWorld, and a visual atlas of agent systems.
Interesting Finds — 2026-09-11 (Evening)
Two more: Tencent's WeKnora knowledge platform and a paper showing sliding-window attention beats linear retrofits.
Interesting Finds — 2026-09-07
Five notes: a one-shot web-demo bench, Cloudflare cache transcoding, Copilot cost efficiency, new knobs for the brain, and Fable 5.1 worlds from code.
Interesting Finds — 2026-09-06
Four notes: Raschka on looped transformers, TxBench antibody evals, Google Pics gated to paid tiers, and the Gemini Flash Cyber overlap.
Interesting Finds — 2026-09-08 (Evening)
Five more notes: Shumer's 3D-world loop, a benchmark that scores agent construction, Qwen-Drive-1.0, loop engineering patterns, and AlphaGenome Atlas.
Interesting Finds — 2026-09-08 (Night)
Eleven more: Images 2.5, Mercury 2.5 diffusion, ID-V2V restyling, cross-model KV transfer, Muse, Gaussian Splat Lite, the KV-caching explainer, tgrep, DIY weather forecasting, a 1-meter LiDAR viewer, and Switzerland's open-source trial.
Interesting Finds — 2026-09-05
Five notes: an SSH fighting game with a bot league, Muse Spark 1.3, the Muse superapp leak, test-time training as a scaling axis, and Meta's organizational second brain.
Interesting Finds — 2026-09-04
Three preprints on harness as infrastructure: HarnessDev measures whether LLMs can build their own scaffolding, Harness-of-Harness makes multi-day autonomy composable, and Fast Weight Attention fixes the temporal alignment of recurrent memory.
Interesting Finds — 2026-09-03
Seven notes: California youth-safety regulation, industrial overcapacity as science catalyst, robotics hardware takeoff, in-context learning's GPT moment, AI as wet-lab co-pilot, a productivity-miracle claim, and Gemini 3.8 Flash / Flash Cyber.
Normalized LoRA: Fixing the Down-Projection Is Enough
A one-line normalization of LoRA's A matrix restores balanced early gradients, faster convergence, and mergeability without extra parameters.
Interesting Finds — 2026-09-02
Seven notes: Qwen Max 0902's 2.4T post-train, Quasar 438B tops Europe, H3-World turns H3 into a world model with 0.2% params, Fable 5.1 decodes a 373-year cipher, Humain M3 in Arabic, a cancer-vaccine primer, plus Dyson's $499 camera-toothbrush — noted with editorial.
A Bench in Golden Gate Park
The same Golden Gate Park Three.js prompt across several models, with the artifact, one-shot result, wall time, and run harness kept together.
Claude Fable 5.1 and Mythos 5.1: Same Weights, Two Safeguard Tiers
25% cheaper cache reads, 60% fewer cyber false positives, and early lab-validated science from a model that ships as two permissions, not two capabilities.
Interesting Finds — 2026-08-27
Five notes: JIT-Agent synthesizes harnesses on the fly, Michigan Robotics open-sources its curriculum, DualView publishes exact phase-specialized weights, plus an ESP32 dashboard and the long-lived free-APIs directory.
Interesting Finds — 2026-08-25 (2/4)
Five more notes: hardware-aware kernels, a tighter matrix multiplication bound, AMD credits, and two SemiAnalysis signals on the CUDA moat.
Interesting Finds — 2026-08-25 (3/4)
Diffusion for language, skill-aware RL, explorative modeling, looped MoEs, and a map of 733k papers.
Interesting Finds — 2026-08-25 (4/4)
Sleep for hybrids, aggressive diffusion decoding, AI-found test-time controllers, a virtual cell, a DeepSeek J-Space report, and PagedAttention explained.
Interesting Finds — 2026-08-25 (5/5)
Meta's Hatch and Watermelon, Perplexity's local-first Portable Computer, Figure's Index dataset, Amazon's automated last mile, Jetson Orin Nano 2, and Keenable's knowledge index.
DiffusionGemma: Parallel Denoising From an Autoregressive Checkpoint
Google DeepMind finetunes Gemma 4 into a discrete diffusion model that generates around 20 tokens per forward pass.
DSpark in vLLM: Verification Budgeted by Confidence, Not by Configuration
One setting keeps speculative decoding fast from 1 to 256 concurrent users by skipping low-confidence drafts when the GPU is busy.
Muse Glimmer 30B: Agents That Stay on the Bench
Meta distilled Spark for the workshop — a 30B model that runs locally and keeps the job on the bench.
Coding Agents Are Not Chatbots
13.5 million GitHub Copilot sessions show what happens after you press enter.
If Intelligence Becomes Cheap, Who Maintains the Result?
Hall proposes AI that makes democracy cheaper to run. The question is who keeps it running.
WeatherNext 2: Forecasting as Ensemble Perturbation
Google DeepMind open sourced its most accurate weather model. The mechanism is simpler than it looks.
