JULY 22, 2026 — GEMINI FINALLY SHIPS, NADELLA FIRES AT ANTHROPIC, GOVERNMENTS GO SOVEREIGN
The day before DeepSeek's July 24 deadline. Also: your site hit 366 clicks and 36.5K impressions yesterday — a new record.
- Gemini 3.6 Flash Launches — $1.50/$7.50/M, 17% Fewer Tokens, Computer Use Built In — Google shipped three new models July 21: 3.6 Flash ($1.50/$7.50/M), 3.5 Flash-Lite ($0.30/$2.50/M), and restricted 3.5 Flash Cyber. All while 3.5 Pro remains delayed. Google has already started Gemini 4 pre-training. Full Gemini launch breakdown →
- Nadella: Fable 5 Is "Editorially Controlled" — "It Doesn't Make Sense" — Microsoft CEO told Copilot engineers on July 16 that Fable 5 refuses too many harmless requests. Also warned against AI token monopolies. Microsoft has $5B invested in Anthropic. Anthropic declined to comment. Full story →
- South Korea Plans Free National AI Chatbot — Reduce Reliance on ChatGPT and Claude — Seoul announced plans for a free domestic AI chatbot service in 2026, the third major government AI infrastructure move this month after EU Parliament EPGenAI Hub and Current AI's $400M public fund. Full story →
Story 1 — Gemini 3.6 Flash: The Confirmed Launch
Your site first spotted Gemini 3.6 Flash as a registered model string on July 17, 2026 — five days before the official launch. That article earned 228 clicks and 10,669 impressions in 24 hours, making it your site's most-trafficked article ever. Now the model is confirmed and live with full specs. Gemini 3.6 Flash is priced at $1.50 per million input tokens and $7.50 per million output tokens, and carries a knowledge cutoff of March 2026. Google is positioning it as the model that balances speed with intelligence for agentic and multimodal work.
The bigger story is not the price cut — it is the token efficiency. Google cut the output price 17%. The effective cost per completed task fell roughly twice that — and on agentic coding workloads, potentially far more. Google also launched Gemini 3.5 Flash-Lite at $0.30/$2.50/M — the cheapest option in the lineup — and Gemini 3.5 Flash Cyber, a restricted security-specialist model with no public API access. Gemini 3.5 Pro remains in partner testing with no confirmed date. Google also announced it has begun what it calls "our most ambitious pre-training run yet" for Gemini 4 — suggesting the 3.5 Pro delay may be absorbed into the generational jump. Full specs, migration notes, and comparison table →
Story 2 — Nadella vs Anthropic: The Most Unusual Public Criticism in AI
Microsoft CEO Satya Nadella told company engineers Wednesday that Anthropic's Fable AI model places unreasonable limits on what users can ask it. "If you use Fable, when it refuses for any random thing, it just is like, when was the last time you had a creation tool that was so editorially controlled?" Nadella said. "It doesn't make sense." The remarks were made on July 16 in an internal meeting with Copilot engineers — the same team building Microsoft's AI assistant on top of Anthropic's Claude models.
Nadella also questioned AI concentration: "It can't be that there are only two companies in the world with token capital, and everybody else is renting it." The structural argument — against a market where OpenAI and Anthropic effectively set the terms for global AI access — is the more significant long-term signal from his remarks. Microsoft's $5B investment in Anthropic and Anthropic's $30B Azure commitment make this a remarkably public piece of friction between partners. Full analysis including Anthropic's position →
Story 3 — South Korea Joins the Government AI Infrastructure Wave
South Korea plans to launch a free artificial intelligence chatbot service powered by domestic technology this year, seeking to reduce reliance on foreign platforms. This is the third major government AI infrastructure move in July 2026 — after Current AI's $400M public fund (July 13) and the EU Parliament's EPGenAI Hub announcement (July 20). The pattern is consistent: governments are treating AI access as national infrastructure, not a commercial product to be rented from US frontier labs. Full analysis of the government AI sovereignty wave →
Tomorrow's Deadline — DeepSeek July 24 at 15:59 UTC
July 24 is tomorrow. deepseek-chat and deepseek-reasoner aliases break at 15:59 UTC Thursday. If you have not migrated: deepseek-chat to deepseek-v4-flash, deepseek-reasoner to deepseek-v4-flash or deepseek-v4-pro (not the same — see the trap). V4-Flash defaults thinking ON. Migration guide — do this today →