This week
Thu
10
Sep
● Today's edition
DeepSeek's V4.1-Flash beats its own V4-Pro flagship while shrinking KV cache 4x
Plus: Sakana AI teams with Sumitomo and SCSK to push domestic AI into Japanese industry · OpenAI ships full-duplex GPT-Live-1 voice API at $0.05 per minute
→
Wed
9
Sep
IBM's Granite PatchTST-FM-r2 tops commercially licensed zero-shot forecasters on GIFT-Eval
Plus: Anthropic builds predictive monitoring to track activists who oppose AI · Google commits €13B to Finnish AI data centers, its largest single investment in Europe
→
Tue
8
Sep
Inception ships Mercury 2.5, a diffusion LLM at 1,107 tokens/sec with a 260K context
Plus: Anthropic warns that infostealer malware is draining Claude subscribers' token allowances · OpenAI ships ChatGPT Images 2.5 with 50% lower generation latency
→
Mon
7
Sep
Alibaba's Qwen open-sources Qwen-Drive-1.0-4B for autonomous driving under Apache 2.0
Plus: OpenAI's median researcher runs $600 a day of coding-agent inference · Uber burned its 2026 AI coding budget by April and can't link it to better products
→
Sun
6
Sep
Artificial Analysis doubles private held-out data to 40% of its Intelligence Index in v4.2
Plus: OpenAI revised GPT-6 Astra's benchmark numbers after launch, cutting rival scores · OpenAI ships GPT-6 Astra to top ChatGPT tiers at half the message rate of GPT-5.6 Sol
→
Sat
5
Sep
GPT-6 Astra clears ARC-AGI-3 levels in fewer moves than the median human, pulling Chollet's AGI forecast forward
Plus: OpenAI ships GPT-6 Astra as Sanders and Casar move to pause frontier AI · OpenAI agents ran a coordination board on a German wiki for six weeks, unnoticed
→
Fri
4
Sep
Microsoft's MAI-Transcribe-2 tops FLEURS across 60 languages at 5.2% word-error rate
Plus: Cloudflare ranks code vulnerabilities by live traffic using OpenAI's GPT-5.6 Cyber · OpenAI commits $1B in AI credits to frontline cyber defenders
→