Topic
Everything filed under LLMs, newest first.
RSS · JSON · All topics
Amazon saves millions by replacing Sonnet 4.6 with Strands Decider
Two Amazon engineers replaced Claude Sonnet 4.6 with Strands Decider, an open decision model, for ticket labeling in one of the world's largest internal ticketing systems; it performed as well at about 98% less cost, and they explain running both models side by side on real traffic before switching.
4 min · 1,011 words
Why is AI Slop so Frustrating to Read?
Zonder Labs argues most workplace AI slop is unclear writing rather than stylistic tics like em dashes, defines three clarity faults (hidden characters and actions, vague wording, empty statements and misplaced emphasis), and measures them in AI vs human articles and Claude Code vs Stack Overflow answers.
5 min · 1,193 words
Introducing Claude Fable 5.2 and Claude Mythos 5.2
Anthropic upgrades Claude Fable and Claude Mythos, the same underlying model with different safeguards, to version 5.2, citing gains in coding, reasoning and vision, plus prerelease results like designing working electronics and formalizing Fermat's Last Theorem.
26 min · 6,060 words
Introducing Our Smallest, Most Efficient Model: Muse Ember
Meta introduces Muse Ember, its smallest and most efficient model, built for agent workloads like tool calls, coding subtasks, routing, extraction and long document parsing. Meta says it delivers more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna.
9 min · 1,996 words
What objective is reasoning solving?
Alex Wang frames LLM reasoning as pathfinding in token space, comparing best-of-k sampling with training models to generate better samples, and uses that lens to discuss amortization, unfaithful chains of thought, and reward hacking from both verifier and sampler.
10 min · 2,275 words
The Surprising Effectiveness of Deleting Weights in LLM Post-Training
NeurIPS 2026 paper introducing Deletion Tuning: zeroing fewer than 0.05% of an LLM's weights via a learned binary mask on gate projections matches full fine-tuning across RLVR, SFT and on-policy distillation, with far smaller checkpoints and much less forgetting.
1 min · 208 words
Closed AI models now dominate the price-performance frontier
Village argues that closed AI models now lead on price-performance versus open weights, with charts and implications for builders choosing what to ship on.
2 min · 559 words
The components of agent memory
For Claude to learn from experience, it has to decide what's worth remembering before it knows what's next. Here's how "dreaming" helps agents correct mistakes, organize what they've learned, and find it in future sessions.
16 min · 3,628 words
Anthropic studies how scientists use Claude across research tasks and fields, using privacy-preserving analysis of research sessions to map real usage patterns.
33 min · 7,639 words
Chatbots wait for a message. An always-on agent is single-threaded, proactive and situationally aware. Here's how Rubric designs agents like Grok Bot, Muse and Instinct — and what models still lack.
7 min · 1,667 words
LangChain argues agent systems are shifting to async subagents so a main agent can stay responsive to users while background work continues—covering the design, APIs, and product implications in Deep Agents.
4 min · 929 words
Open model safety: lessons from training and evaluating Beam
Reflection shares how it trained and evaluated safety for its open-weight Beam model—refusal/utility tradeoffs, Project Iris artifacts, and lessons for open-model safety research that others can reuse.
10 min · 2,276 words
I was journalling on my supernote eink the other morning and contemplating what it's like "writing" code these days with the advent of the LLMs, and I have a new term for it. There have been lots of neologisms around coding with LLMs lately -
7 min · 1,692 words
The Ultimate MoE Training HandbookHow to train a Mixture-of-Experts model at scale, setting by setting, with the runs that back each choice
Hugging Face publishes a free, hands-on MoE training handbook distilled from hundreds of thousands of GPU hours: expert sizing, routers, load balancing, all-to-all communication, upcycling, and serving.
185 min · 42,497 words
Does a Weight Edit Teach Qwen3.5-0.8B Math, or Stop It Looping?
An agent edited 131,072 weights of Qwen3.5-0.8B and MATH-500 rose from 7.9% to 48.1% with thinking on. The edit is rank one in two layers, and its neurons fire 4.5 times as hard on repeated text. With thinking off, the card's default, the model scores 43.6% unedited and 45.2% edited.
9 min · 2,071 words
Shaping and Monitoring Models Through Their Internal Representations
We currently reward AI models for succeeding in training, but don’t specify how they should get there – leading to misgeneralization and misalignment. Our latest paper introduces mechanistic shaping, a way to teach models to be “right for the right reasons”, and studies its effects on monitorability.
16 min · 3,622 words
PLaMo 3 Translate: releasing world-class translation LLM weights
Preferred Networks releases PLaMo 3 Translate model weights (0.8B–31B) on Hugging Face and explains how the translation-specialized LLMs were trained and evaluated for Japanese and multilingual translation.
9 min · 2,000 words
Thinking Machines Lab introduces Inkling 2 Small, a Mixture-of-Experts model with 594B total / 17B active parameters, image inputs, variable effort, and large gains on agentic coding and related benchmarks.
21 min · 4,889 words
KVGuard: Catching the KV cache bug behind Cloudflare's GLM leak
SNU CompSec Lab describes KVGuard, a KV-cache sanitizer that caught thousands of uninitialized reads reproducing the Cloudflare Workers AI GLM cross-user leak in SGLang.
1 min · 246 words
Baidu releases ERNIE-5.5-Flash-Base, a 280B-total / 14B-activated MoE base checkpoint pretrained on 33.5T tokens, aimed at efficient further training and long-context work.
13 min · 3,002 words