Tech • AI • Robotics • Game

ENFR
TodayPlay for freeFury BloxLeaderboardShortsTop StoriesFor youTopicsVideosYT channelsArchivesSearchFavorites

Claude Fable 5.2 launches as OpenAI GPT-6 Cyber finds 948 Linux exploits

AITuesday, November 10, 2026· 28 videos

Briefing

Audio player

Claude Fable 5.2 debuts

Anthropic launched Claude Fable 5.2 and the restricted Claude Mythos 5.2, presenting them as its strongest models for coding, reasoning, vision and long-horizon agent work. The company says the pair is about 30% faster and roughly 20% cheaper than Fable 5.1, with pricing at $8 per million input tokens and $40 per million output tokens. Fable 5.2 is the broadly available version with tighter cyber and biology safeguards, while Mythos 5.2 is limited to verified organizations. Anthropic also claims sharply fewer mistaken refusals and better performance than Opus 5.5, GPT-6 Astra and Gemini 4 Carbon on many internal comparisons.

Fable reaches into hardware

Beyond software, Anthropic showcased Fable 5.2 designing complete electronic prototypes with partner Diode. From short prompts, the model reportedly produced specifications, schematics, PCB layouts, enclosures and firmware for a USB conference speaker, a home voice assistant and a virtual pet in less than one day. In one case it recovered from a failed routing tool by generating a replacement in about 30 minutes and continuing the job. The demo suggests AI’s economic value in engineering may shift from drafting and layout toward testing, certification and final specification.

OpenAI sharpens GPT-6 Cyber

OpenAI detailed a specialized offensive-security system, GPT-6 Cyber, built to find, exploit and help remediate software flaws. In the latest Linux development branch, the company said a multi-agent pipeline uncovered 948 unique local privilege-escalation exploits and generated patches, while broader testing confirmed 1,724 real issues and 1,169 working root exploits. Separate results cited 81 serious unknown flaws in Firefox and hundreds of previously unknown Linux weaknesses. Access is being kept restricted, underscoring how quickly frontier models are lowering the cost and speed of vulnerability research.

Google unveils Gemini 4 Carbon

Google introduced Gemini 4 Carbon as a frontier successor to Gemini 4 Argon, aimed at cybersecurity and long multi-step tasks. Early demonstrations emphasized cleaner coding output, better bug fixing and stronger writing, including more polished interactive projects generated inside Google Anti-Gravity. Compared with Argon, Carbon was described as more playable, visually refined and more reliable at following instructions in code-heavy creative tasks. The release adds another top-tier competitor just as Anthropic and OpenAI push more specialized and higher-agency systems.

OpenAI stealth update rumors grow

Reports from advanced users suggest some GPT-6.1 Soul coding sessions inside Codex may be getting routed to a newer checkpoint informally dubbed GPT-6.2 Soul. The strongest claims center on noticeable jumps in coding quality, tool use and visual generation, especially when the model is run at High reasoning settings. Informal knowledge-cutoff probes, including questions about the Tony Awards on June 7, 2026, produced mixed and inconclusive evidence. With no official announcement from OpenAI, the episode highlights how opaque model routing and silent updates have become for power users.

Nvidia rocked by Jensen Huang scandal

Nvidia faced a rare leadership crisis after reports that Jensen Huang acknowledged a relationship with Alice Wong or Alice Wang, a business figure linked to Mega Speed, an Asia-based AI infrastructure reseller under scrutiny. Reporting tied the case to possible restricted-chip diversion and suspected links to Chinese intelligence, turning a personal matter into a governance and national-security issue. Nvidia shares were said to fall between 5% and 8% as investors weighed potential damage to Huang’s standing in Washington and to the company’s compliance narrative. The controversy lands at a sensitive moment for the global AI chip trade, where Nvidia is central to export-control enforcement.

AI infrastructure selloff hits debt plays

The recent plunge in AI infrastructure stocks appears to reflect a financing squeeze more than a collapse in compute demand. Between October 8 and November 6, shares of CoreWeave, Nebius and IREN fell roughly 40% to 50%, while Microsoft, Amazon, Meta and Alphabet were far more resilient. Credit stress surfaced earlier, with a CoreWeave bond issued in June at just over 10% yielding around 15% by October 13, signaling rising default-risk concerns. The split reinforces a two-tier market in which hyperscalers can fund expansion from cash flow while debt-heavy neo-clouds remain exposed to refinancing shocks.

Polymarket, Binance and Ledger breached

A string of crypto security incidents widened concern about operational controls across financial platforms. Polymarket was hit by a front-end supply-chain compromise after attackers reportedly altered DNS settings tied to DebugBear and swapped in wallet-draining JavaScript through a CDN script. Reimbursement terms reportedly required some users seeking $10,000 or more to complete identity checks and sign a non-disparagement clause, drawing criticism. In the most consequential case, the FBI linked an alleged $6 billion theft involving Binance to North Korean actors, underscoring the scale of state-backed cybercrime.

Videos covered

Previous briefings · AI