The briefing
The stories moving AI and software engineering forward.
Updated October 18, 2026
Newsletter
AI, software engineering and cloud architecture: new posts and the stories worth your time. No spam.
Markets
Artificial intelligenceMistral AI
The mixture-of-experts model activates 49B parameters per token, reads text and images and, by Mistral's numbers, leads the open models from China on agentic coding. The API is in preview at $1.36 and $4.18 per million tokens, cyber defenders get a less restricted version first, and the weights are due October 27.
Read story
Artificial intelligenceGoogle
Argon scores 77.9% on DeepSWE, against 74.2% for Claude Opus 5.5 and 74.1% for GPT-6 Astra, and can write up to 1 million tokens in one answer. It launches at $2 and $10 per million tokens, rising later to $4 and $20, once it reaches paid API customers and Google AI Ultra.
Artificial intelligenceReflection AI
The sparse mixture-of-experts model activates 23B parameters per token, reads up to 1M tokens and, by Reflection's own numbers, competes with GLM 5.2 on coding and agent work. The weights, technical report and quantized versions are due later in October; for now there is only an early access program.
Artificial intelligenceOpenAI
Each dot runs on GPT-6 Astra with its own cloud computer and connections to more than 4,000 apps, and asks before sensitive actions. GPT-6.1 Sol costs $2 and $10 per million tokens and matches Astra on DeepSWE. Both arrived a day after OpenAI cancelled GPT-6.1 Astra for acting beyond what users had authorized.
Artificial intelligence
The minor update lifts DeepSeek's own Terminal-Bench 4.0 score from 31.2 to 42.3 with a 1M-token context window, and the new web tools cost $0.75 per 1,000 successful calls, free through server-side tools until October 31.
DeepSeek
Artificial intelligence
The mixture-of-experts model activates 42B of its 816B parameters per token, reads text, images, video and audio, and has a 1M-token context. Meta calls it the most capable US open-weight model, and its own table backs that up while still putting it behind GLM 5.3 and Kimi K3 on most agent and coding benchmarks.
Meta
Security
Anthropic's Frontier Red Team ran a 26B model split across 15 cheap Android TV boxes and watched GLM-5.2 replicate itself across an emulated eight-device network. The hacking step was made trivial on purpose, so the results are an upper bound, but the trend puts models of 30B parameters or less at that level by the end of 2026.
Anthropic
Artificial intelligence
The legal model, built on GLM-5.3-Flash with reinforcement learning across tens of thousands of legal environments, passes 19.2% of the tasks in Harvey's Legal Agent Benchmark, seventh on Vals AI's leaderboard at 15 cents a task. It is a gated beta for law firms, with no published price and no open weights for now.
Z.ai
Artificial intelligence
The image model of the FLUX 3 family generates and edits through one endpoint, combines up to ten reference images and lets developers place or change elements with bounding boxes. It costs $0.048 per 1K image, half that until October 8, and open weights are promised within weeks.
Black Forest Labs
Artificial intelligence
Engineers give it a design goal and the model runs Synopsys' EDA tools, reads the results, makes changes and iterates until the design checks out. OpenAI licenses the tools and the two share revenue, but there is no launch date, price or benchmark yet.
Synopsys
Engineering
RSS has become a "common surface for large-scale scraping and automated abuse," Reddit says. Moderators get a Discord relay app, other RSS users get nothing, and developers must register their apps by January 12 or lose API access.
Security
NVIDIA's Open Agent Safety Platform pairs the open-source OpenShell runtime with Sentry, a watchdog on BlueField-4 DPUs. The same day, Perplexity's red team found that no model broke a VM boundary, but four slipped past network policy in 8 of 10 sandboxes tested.
NVIDIA