Pulse Daily — Sunday, August 16, 2026

On Sunday, August 16, 2026, the AI Pulse covers a week of transformative AI developments: OpenAI's unreleased Astra model cracked ten unsolved math problems for $2,000 in compute; ChatGPT crossed one billion users while GPT-5.6 Luna prices fell 80%; Meta's open-weight Muse Glimmer arrived for on-device agentic work; Google unveiled HEIR homomorphic encryption for private AI inference; Alibaba's Qwen 3.8 27B pushed open-source frontiers; DARPA flew an AI-piloted F-16 without a human at the controls; INTERPOL found AI driving 55% of cybercrime in Africa with losses doubling to $484 million; OpenAI's IPO S-1 filing neared; the tl;dv AI notetaker exposed 181,000 meeting records in a major privacy breach; OpenAI's Daybreak initiative expanded AI safety access tiers; and frontier labs publicly declared the early stages of an intelligence explosion. Stories: Mathematicians Begin Scrutinizing Astra's Proofs — Verification Meets Reality · GPT-5.6 Luna Is Now ChatGPT's Free Default — Frontier Reasoning at Zero Cost · OpenAI's S-1 Expected Any Day — First Full Financial Disclosure Before IPO · Meta's Muse Glimmer Puts a 30B Agentic Model on Consumer Hardware · Qwen 3.8-27B Lands on Hugging Face — Vision, 262K Context, Apache 2.0 · Google's HEIR Compiler Runs AI Models on Fully Encrypted Data — No Plaintext Needed · DARPA's AI-Piloted F-16 Completes Full Autonomous Flight Test Program · Interpol: AI Powers 55% of Cybercrimes Across Africa — Losses Double to $484M · tl;dv's AI Notetaker Exposed 181K Meeting Records for Six Months · GPT-5.6-Cyber Finds Two Chrome Zero-Days — Daybreak Red Is Live.