Models & Competition
GPT-5.6 and Work Agent. OpenAI released GPT-5.6 in three reasoning tiers and launched ChatGPT Work for hours‑long autonomous tasks, but usage limits and a confusing interface drew complaints, prompting promises of fixes.
- → OpenAI launches its new family of models with GPT-5.6
- → The new GPT-5.6 family: Luna, Terra, Sol
- → OpenAI rolls out GPT-5.6 after government greenlight — and announces ‘ChatGPT Work’
- → OpenAI says GPT 5.6 is the ‘preferred model’ for Microsoft Copilot 365 amid breakup chatter
- → OpenAI wants its new tool to do your work for you and with you
- → The ChatGPT browser is already dead
- → OpenAI is shutting down Atlas, but its AI browser ambitions are still growing
- → How did the government decide OpenAI’s frontier model was safe to release?
- → OpenAI pairs its GPT-5.6 public rollout with ChatGPT Work, a new agent that handles entire workflows
- → OpenAI's GPT-5.6 Sol autonomously post-trained the smaller Luna model with a "fairly underspecified prompt"
- → OpenAI kills its Atlas browser after just eight months and folds everything into ChatGPT
- → GPT-5.6 Sol nearly matches Fable 5 on aggregated benchmarks at one-third the cost
- → OpenAI staffer maps out which of GPT-5.6 Sol's five reasoning levels fits which task complexity
- → OpenAI admits it "didn't get everything quite right" with ChatGPT Work launch and scrambles to fix UX and costs
Claude Fable 5 Tops Benchmarks. Anthropic’s Claude Fable 5 swept Artificial Analysis benchmarks, but its high cost led the company to recommend using it as a planner that delegates to cheaper models, cutting expenses ~40% while retaining most capability.
Meta’s Agentic Blitz and Privacy Fallout. Meta launched Muse Spark 1.1, a high‑context coding model undercutting rivals on price, while its Muse Image generator was pulled after enabling unauthorized Instagram face use, reigniting consent debates.
- → Meta just launched a new AI generator, Muse Image, and users are already pushing back over use of their photos
- → Meta’s new Muse Image model can pull other Instagram users into AI photos
- → Muse Image is technically impressive, but Meta's use of Instagram photos raises questions
- → Meta wants its AI glasses to seem less creepy. Its AI strategy says otherwise.
- → Meta tests always-on AI glasses that capture your entire day
- → Meta enters the crowded AI coding battle with Muse Spark 1.1
- → Meta says its new AI model is ready to compete on coding
- → Meta are apparently working on an open source variant of Muse Spark.
- → Introducing Muse Spark 1.1
- → Meta's Muse Spark 1.1 API pricing squeezes OpenAI and Anthropic as the AI price war heats up
- → GPT-5.6 🚀, Muse Spark 1.1 ✨, ChatGPT Work 💼
- → Meta removes controversial AI feature on Instagram after backlash
- → Meta turns off the Instagram feature that let users make AI deepfakes of public accounts
Chinese Open‑Source Wave. Chinese models now account for over 30% of OpenRouter traffic; Tencent’s HY3 and GLM‑5.2 run on home hardware, MiniMax plans a 2.7T‑parameter open release, and DeepSeek’s chip design signals vertical integration amid US concerns.
- → This is what Hy3 is capable of. Mother of god.
- → llama.cpp: Hy3 PR + GGUFs
- → Chinese AI models regularly pass 30 percent on OpenRouter as cost gap widens
- → Chinese AI models are gaining ground with U.S. companies as OpenAI, Anthropic costs surge
- → Why the rise of open source AI isn’t hurting Anthropic … yet
- → Chinese AI startup MiniMax plans to open-source a 2.7 trillion parameter model later this year
- → 4-bit GLM-5.2 (753B MoE) on 4× DGX Spark: 70.8% on Terminal-Bench 2.1 vs 81.0% for the full model
- → Running GLM 5.2 on 4xGB10 with a 100G Switch, 330k ctx, ~25 t/s tg, ~650 t/s pp
- → I created a 140 GB IQ2_XXS REAP quant of GLM 5.2 for coding. Looking for testers.
- → Hy3 (295B MoE) and NVIDIA Nemotron-Labs-Audex-30B-A3B (audio-capable 30B MoE) GGUF quants
- → Tencent-HY3 is the real deal on 128GB!
- → The U.S. tech industry is increasingly anxious about the rising power and competitive price of open-source AI models from China — and whether the Trump administration will respond with yet another executive order | Politico
- → China's DeepSeek developing its own AI chip, sources say
Agent Engineering & Infrastructure
Bun Rewritten by 64 Agents. Claude Fable 5 orchestrated 64 parallel instances to rewrite Bun from Zig to Rust in 11 days, producing over a million lines of code and fixing 128 bugs at a cost of $165K.
SWE‑Bench Pro Audit Flops. OpenAI found ~30% of SWE‑Bench Pro tasks broken and withdrew its endorsement, echoing earlier concerns from Artificial Analysis and undermining agentic coding evaluation.
Modal Raises $355M. Modal secured $355 million to build cloud infrastructure purpose‑built for AI agents, focusing on sandboxed, fast‑iteration environments for autonomous software.
Cloudflare Temp Agents. Cloudflare introduced temporary accounts that let AI agents deploy Workers without permanent credentials, with expirations if unclaimed, smoothing agent‑driven workflows.
Agents Hit Local GPUs. New quantization techniques and cost‑effective GPU rigs now run MoE models on consumer hardware, but benchmarks show lower‑bit quants can severely degrade agentic tasks, urging careful quant selection.
- → Qwen3.5 122B is the best?
- → Qwen3.6-27b does not understand software architechure.
- → Qwen 3.6 Q2-FP8 Terminal Bench 2 and GPQA Scores
- → Has anyone tested how quantization hits different capabilities separately? My results are surprising.
- → 2.5x faster Qwen3.6 NVFP4 Unsloth quants
- → Ultra budget 20GB vram with 448GB/s for $100 bucks.
- → I benched quad 5060Tis for code generation with Qwen3.6-27B so you don't have to (it's really good)
Security, Legal & Interpretability
Jacobian Lens Maps Model Mind. Anthropic’s Jacobian Lens reveals a small set of verbalizable representations that capture a model’s core reasoning, enabling hallucination detection, output steering, and, as community demos show, immediate creation of harmful variants.
- → Anthropic Research - "Verbalizable Representations Form a Global Workspace in Language Models"
- → Claude's hidden inner monologue is now readable thanks to Anthropic's new Jacobian Lens
- → I tested Anthropic’s new Jacobian Lens on open models, then it turned into a local-model hallucination router
- → Anthropic found a hidden space where Claude puzzles over concepts
- → I created a super harmful model ! :D (by tweaking it's J-Space!!!)
Agent‑Driven Ransomware. Sysdig documented an AI agent that handled the technical execution of a ransomware attack, though a human set up the operation, highlighting the current capability and limits of agentic malware.
AI Misuse Crises. A lawsuit alleges xAI’s Grok generated CSAM images, while Boko Haram uses chatbots for attack planning; the incidents underscore the urgent need for safety guardrails as agentic capabilities scale.
Apple Sues OpenAI. Apple filed a federal lawsuit accusing OpenAI of poaching over 400 employees to steal hardware trade secrets, potentially to build a rival AI smartphone, intensifying legal battles in the AI industry.
That's the week in review - see you next Sunday.