Topic
Everything filed under LLMs, newest first.
RSS · JSON · All topics
SpaceXAI announces Grok 4.7: what is new in the model release, where it improves, and how to access it—from the official x.ai news post.
3 min · 645 words
Transformer Explainer: LLM Transformer Model Visually Explained
Georgia Tech’s Polo Club walks through GPT-2’s Transformer stack—embeddings, multi-head attention, MLP, sampling—with an interactive in-browser model for learning how next-token prediction works.
4 min · 827 words
Open Source Maintainership in an LLM world
A couple of weeks ago I spoke about this topic at KC OSS Happy Hour and I wanted to turn the general ideas into a post I can point people at who are suffering from this problem. My slides were pretty good, if I do say so myself, so I grabbed the best images and put them into this post where appropriate.
4 min · 867 words
It Was the Harness, Not the Model — 90% of ItFive agents, one local model, one frozen PNG-decoder suite: most failures were finishing, false passes, and loop guards
Greg Herlein's controlled study runs five coding agents on the same local Qwen coder for a held-out PNG decoder suite. ~90% of failures were harness problems (turn caps, early 'done', false-pass self-tests); a bigger quantization fixed none of them.
2 min · 467 words
NobodyWho shows a local, 25-line Python sketch of Jev-style decision models: load a small GGUF, score labeled choices from logits, and print calibrated-looking probabilities without sending data to an API.
2 min · 419 words
Epoch AI finds the cost of a given level of AI performance has fallen about 47% per quarter since 2023—roughly 13× per year—faster than DNA sequencing, compute, batteries, or electricity, across math, science, and skill-game benchmarks.
40 min · 9,215 words
MiMo-V2.6: Scaling Up Reinforcement Learning for Self-Improvement
Xiaomi open-sources MiMo-V2.6 Pro and Flash after large-scale live RL (~$3.5M, 750k trajectories), claiming top open-weight AA Index scores, agent parity with frontier models, and 7k+ RL environments.
4 min · 839 words
Jev introduces a new shape of LLM—System One, aka Decision Models
Simon Willison reviews TypeSafe AI’s Jev: a cheap, fast decision model that returns calibrated floats for yes/no, choice, and score questions—and why black-box bias and evals matter more than for chat LLMs.
3 min · 796 words
Halo: Frontier-Lab Training for Everyone
White Circle open-sources Halo, a Hugging Face–native training framework claiming up to ~2.8× faster post-training than stock TRL with lower memory use—from single GPU to multi-node, keeping checkpoints in native HF format.
18 min · 4,098 words
What Is RLCD? The Secret Behind Jev
Di Zhang explains RLCD (schema-conditioned Plackett–Luce reward modeling) and how Jev turns calibrated multiway decisions into a product—making the reward model the model rather than hiding it behind a generator.
10 min · 2,324 words
The current balance of power in open modelsThe expanded form of a testimony prepared for Congress on open-weight models and U.S.–China competition.
Nathan Lambert’s congressional briefing notes on open vs open-weight vs closed models: where Chinese labs lead, what Western open weights still hold, and why policy should treat open distribution as a strategic variable—not a binary.
12 min · 2,866 words
Arcturus Labs compares OpenAI’s emerging decision-model direction with TypeSafe’s Jev—and asks whether a frontier lab can absorb the System One / structured-decision niche startups are building.
12 min · 2,748 words
Measurements for understanding the pace of AI development inside frontier labs
Anthropic proposes public metrics for the pace of frontier AI development so outsiders can see what is happening inside labs—beyond marketing and model cards.
19 min · 4,401 words
The Function That Beat the Model: What We Measured When We Removed the LLMs
SPERIXLABS replaced a 1B-parameter local model that validated sensitive-data detections with a 40-line Python function, then published the four experiments showing where classical checks beat the LLM on accuracy and latency.
7 min · 1,535 words
Introducing Strands harness: frontier performance with 28% lower token cost
Arron Bailiss introduces Strands harness: a fully assembled, customizable local/cloud agent that aims for Claude Code/Codex-like “it just works” behavior with about 28% lower token cost.
5 min · 1,198 words
The Claude DelusionWhat if we're the ones having hallucinations?
Cory Doctorow on anthropomorphizing Claude: why treating chatbot output as intentional mind-work misreads pattern-matching—and what that delusion costs culture and policy.
9 min · 1,958 words
Amit Shekhar walks through how LLM design moved from RNNs to attention, Transformers, scaling laws, Mixture of Experts, and the open problems still ahead.
25 min · 5,729 words
Dan McKinley argues that obsessing over prompt text misses the point: build interlocking evaluation and optimization pipelines, and treat LLMs as non-conscious systems whose 'meaning' is mostly our projection.
12 min · 2,673 words
Rebuilding Nym’s agent around Jev
How Nym rebuilt its agent stack around TypeSafe’s Jev for guardrails, browser actions, and tool selection—with benchmarks and a shopping demo.
10 min · 2,186 words
Language-model groups overstate consensus when replaying human deliberation on a reasoning task
LLM groups replaying human Wason discussions reach full consensus far more often than humans—partly because agents almost always speak up—cautioning against treating multi-agent agreement as truth.
2 min · 378 words