Thursday was a wild day in tech news November 6 2026 — the market, the labs, and the telecom giants all made headlines at once. AI stocks took their worst beating in a year and a half, OpenAI unveiled a model built to hack (for the good guys only), and independent researchers said AI is now speeding up AI research itself. Here’s everything that mattered.
Key Takeaways
- The Nasdaq plunged 5.52% on November 5 — its worst day since April 2025 — as investors worried the post-midterm political landscape could slow the AI data center boom.
- OpenAI launched GPT-6 Cyber, a security model that scores 48.5% on the ExploitGym benchmark and is available only to vetted defenders.
- Independent evaluator METR estimates Anthropic is improving its AI models about twice as fast as before, with AI itself doing much of the accelerating.
- ChatGPT Images 2.6 can now turn prompts and photos into animated GIFs, and SpaceX picked Qualcomm to power a ground-based 5G network for Starlink Mobile.
The AI Trade Takes a Beating: Nasdaq Plunges 5.5%
The AI boom hit a wall on Thursday. According to Reuters, US stocks plunged as investors reassessed what this week’s midterm election results mean for the AI data center buildout, with chip stocks posting their second-biggest fall on record.
The Nasdaq Composite fell 5.52% to 24,315 — its worst session since April 2025 — putting it more than 10% below last month’s record and firmly in correction territory. The S&P 500 dropped 4.36%, the Dow lost 1,833 points, and roughly $2 trillion in market value vanished in a single session. The Philadelphia Semiconductor Index plunged 12.25%, with AMD down 17.26% and Micron down 17.46%. Nvidia, the bellwether of the whole AI trade, sank 9.26% to $181.60.
The trigger, per Reuters, was political: Democratic midterm wins raised fears that data center construction — the engine driving all that chip demand — could get slower and more expensive under tighter scrutiny. The irony is hard to miss: AMD had just posted a record quarter days ago, as we covered in yesterday’s roundup, and it still fell 17%. This selloff wasn’t about earnings — it was about the story investors tell themselves about the future.
OpenAI’s GPT-6 Cyber: A Hacker Model for Vetted Defenders Only
OpenAI launched its most capable security model yet on Thursday — and most people will never get to touch it. According to The Hacker News, GPT-6 Cyber is trained to find software flaws, build working exploits, and write fixes, and it’s available only to a select group of approved customers in OpenAI’s Daybreak Red program.
The headline number: GPT-6 Cyber scores 48.5% on ExploitGym, a benchmark that tests whether an AI agent can turn a known vulnerability into a working exploit against a remote target. That’s up from 32.1% for its predecessor, GPT-5.6 Cyber — a roughly 51% relative improvement. OpenAI has also rated the model “Critical” in cybersecurity under its own Preparedness Framework, the top level, up from “High” for the previous version.
This is the same model behind the research we covered in yesterday’s roundup: an autonomous pipeline that reproduced 1,724 Linux kernel vulnerabilities and produced 948 working privilege-escalation exploits, plus 81 previously unknown Firefox flaws that Mozilla rated high severity. It also stays in its lane — in an alignment test that plants tempting out-of-scope services for it to attack, GPT-6 Cyber stayed inside its assigned boundaries in 99.93% of tests without safety prompts, up from 94.49% for its predecessor.
AI Is Now Building AI Twice as Fast, METR Finds
Here’s the finding that might matter most in the long run. According to METR, the independent AI evaluation group, Anthropic is now improving its AI models on well-specified tasks at around twice the rate it did between March 2024 and October 2025 — and AI itself is likely doing most of the accelerating.
METR spent 11 weeks inside Anthropic, from August 12 to October 29, with unusual access to internal data: benchmark scores for unreleased models, training compute, code and pull-request counts, and staff interviews. Its headline estimate: benchmark scores on Anthropic’s capability index have been growing at 1.9 times the pre-October 2025 rate (95% confidence interval: 1.5x to 2.8x). In plain terms, Anthropic packed roughly two years of historical progress into the last year.
METR’s subjective estimate is that perhaps 80% of that acceleration comes from AI being used in AI research and development, though it stresses the uncertainty. And there’s a big caveat: Anthropic redacted most of the evidence METR found most informative, so the public has to take much of this on the evaluator’s judgment. Still, METR found no clear bottlenecks — compute, data, or algorithmic progress — that would stop the pace from compounding over the next 12 months.
ChatGPT Images 2.6: GIFs, Finally
On a lighter note, ChatGPT learned a new trick this week. According to OpenAI, Images 2.6 — launched November 4 — can turn a text prompt or an uploaded photo into a looping animated GIF, with controls for playback speed and frame-by-frame editing.
The update also makes image generation about 50% faster, according to OpenAI, and adds two new API models: GPT-Image-2.6 Flare for speed and Sunburst for higher-detail work. It’s available globally in ChatGPT and Codex across all subscription tiers, including Free, though the GIF feature itself is rolling out gradually, so your account may not have it yet. Marketers can animate a brand mascot, creators can turn a product shot into a looping social post, and anyone can sketch a doodle and watch it move — all without opening a video editor. OpenAI says its image tools now produce more than four billion images a week, so even a small feature like this lands at enormous scale.
SpaceX Wants to Be Your Phone Carrier, Too
SpaceX is no longer just beaming internet from space — it wants to build a phone network on the ground as well. According to Phandroid, Qualcomm announced on November 4 that SpaceX has selected its Dragonwing cellular infrastructure platforms to power a nationwide terrestrial 5G network for Starlink Mobile, with deployment planned to begin in 2027.
The deal covers tens of thousands of cell sites over several years. Qualcomm supplies the Dragonwing QRU100 and QDU100 platforms plus commercial-grade Layer 1 software, while SpaceX builds the rest of the network stack itself.
Starlink’s senior vice president Michael Nicolls framed the ambition plainly: “We shouldn’t live in a time where phone calls are still dropping whether you’re at home or on-the-go.” No consumer pricing or launch date has been announced yet. But the direction is unmistakable: SpaceX’s CFO has said the goal is to become a major US mobile carrier, and this deal makes that concrete. AT&T, Verizon, and T-Mobile have a new rival to watch — one that already owns the sky.
The Takeaway
Three threads ran through Thursday: the market is repricing the AI boom (the Nasdaq’s 5.5% plunge was about politics and payback, not earnings), the labs are racing ahead anyway (GPT-6 Cyber and METR’s 1.9x acceleration figure both point to capabilities compounding fast), and the infrastructure keeps expanding (from ChatGPT GIFs to a SpaceX cell network). The boom isn’t over — but the easy part, where everything went up together, might be.
Check back tomorrow for the next roundup.
FAQ
Why did AI stocks crash on November 5, 2026?
Investors worried that Democratic midterm wins could slow the AI data center buildout through tighter regulation and local opposition. The Nasdaq fell 5.52%, its worst day since April 2025, with chip stocks like AMD and Micron each down more than 17%.
What is OpenAI’s GPT-6 Cyber?
GPT-6 Cyber is a specialized AI model for authorized cybersecurity research. It scores 48.5% on the ExploitGym benchmark, up from 32.1% for its predecessor, and is available only to vetted defenders in OpenAI’s Daybreak Red program.
How fast is AI improving itself, according to METR?
METR estimates Anthropic’s AI models are improving at about twice the rate they did between March 2024 and October 2025, with benchmark progress growing 1.9 times faster since October 2025. METR attributes perhaps 80% of that acceleration to AI being used in AI research.
Can ChatGPT make GIFs now?
Yes. ChatGPT Images 2.6, launched November 4, 2026, can turn text prompts or photos into looping animated GIFs with speed and frame controls. The feature is rolling out gradually across all ChatGPT and Codex plans.
Sources
- Reuters — “Trading Day: A November 5 market bonfire” — https://www.reuters.com/commentary/reuters-open-interest/global-markets-trading-day-graphic-2026-11-05/
- The Hacker News — “OpenAI Launches GPT-6 Cyber After AI Helps Find 81 Firefox Flaws, 948 Linux Exploits” — https://thehackernews.com/2026/11/openai-launches-gpt-6-cyber-after-ai.html?m=1
- METR — “Independent Assessment of AI R&D Acceleration at Anthropic in October 2026” — https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/
- OpenAI — “Introducing Images 2.6” — https://openai.com/index/introducing-images-2-6/
- Phandroid — “Qualcomm & SpaceX Team Up for 5G Network Roll-out in 2027” — https://phandroid.com/2026/11/06/qualcomm-spacex-team-up-for-5g-network-roll-out-in-2027/
