170,000 signatures can't bring back Jensen Huang - or can they?
8 min read · 15 sources
- A petition for Jensen Huang's reinstatement as Nvidia CEO surpassed 170,000 signatures, backed by Musk, Tan, and Sweeney, but a national security investigation blocks his return.
- SpaceX caught its full Starship in midair at Starbase, Texas, after an orbital flight, proving the reusable launch system concept.
- Tesla opened Cybercab pre-orders at $29,990 with a $299 monthly operating fee, geofenced to Texas cities with December 2026 delivery.
- Marathon Fusion claims it made gold from mercury, a transmutation requiring roughly a million times more energy than chemical bonds, though quantities remain unpublished.
- OpenAI and Anthropic are earning over $115 billion annualized, with revenue tripling yearly for four years, suggesting digital-worker AGI in 2-4 years.
The most powerful man in AI chips is out of a job, and the tech world is throwing a fit. A petition demanding Jensen Huang’s reinstatement as Nvidia CEO has blown past 170,000 signatures, with Elon Musk, Y Combinator’s Garry Tan, and Epic’s Tim Sweeney all publicly backing it. But here’s the thing: Huang resigned amid reports of extramarital affairs, one involving a woman suspected of ties to Chinese intelligence, and the board is investigating how his actions damaged America’s AI dominance. CFO Colette Kress is running the show in the interim.
For engineers, this isn’t just celebrity gossip. Nvidia is the sole supplier of the GPUs that train every frontier model, and leadership instability at the top means uncertain roadmaps, potential export policy shifts, and a board that’s suddenly very sensitive to national security optics. The investigation into Huang’s relationship with President Trump and his Cabinet makes a return unlikely, no matter how many signatures pile up.
AIs as good as the best humans at AI research are only months to a couple of years away, potentially leading to superintelligence by 2030.
Nvidia's Leadership Vacuum: What a Huang Return Would Actually Change
The petition has reportedly attracted support from Tesla CEO Elon Musk, Epic Games CEO Tim Sweeney, and Y Combinator CEO Garry Tan, per CNBC. Huang has discussed the possibility of returning, but the investigations surrounding him will likely prevent that from happening.
The board is investigating how his actions may have damaged America’s AI dominance, the national security implications, and his close relationship with President Trump and his Cabinet. For anyone running AI workloads, the practical question is whether Kress can hold the line on GPU allocation, pricing, and the Blackwell roadmap while the board searches for a successor. A prolonged vacuum at Nvidia’s top doesn’t change what chips ship this quarter, but it does change how customers negotiate multi-year supply agreements.
SpaceX Catches Starship in Midair: The Reusability Bet Pays Off
SpaceX successfully caught its Starship rocket in midair at its Starbase facility in South Texas on Saturday, after the vehicle returned from orbit, per Bloomberg. The vehicle descended belly-first, then flipped upright to land next to the launch tower, with the tower’s mechanical arms catching it.
This is the first full recovery of an orbital-class vehicle, and SpaceX has now captured both sections of the launch system during different missions. It aborted an attempt to catch Super Heavy during the recent flight due to a slight underperformance of the engines. For anyone tracking launch economics, this is the difference between throwing away a 747 after every flight and reusing it. Full reusability is the only path to the launch costs that make mega-constellations and deep-space missions financially sane.
Apple Bets Its Smart Glasses on Siri AI Apps
Apple’s upcoming smart glasses, planned for 2027, will center on Siri AI integrations as their main draw, according to a Bloomberg newsletter. The company is counting on integration with apps people already use on their iPhones as a key differentiator, betting its large developer ecosystem gives it an edge over Meta and Google.
The hardware won’t be dramatically different from rival offerings, which makes the developer ecosystem crucial. Apple also needs software makers to begin preparing now. The article notes that bringing Face ID to the iPhone Duo is a top priority, but Apple is running out of time to include it in the second version. For developers, this is the classic Apple play: ship the hardware, win on the platform. The glasses are a Siri delivery vehicle, not a hardware story.
One Proton Away: Marathon Fusion Claims Gold from Mercury
Marathon Fusion, a San Francisco-based fusion startup, claims it has created gold from mercury, per Konvoy. Changing one proton - gold’s 79 to mercury’s 80 - requires roughly a million times more energy than breaking a chemical bond. The startup hasn’t published how much gold it made, and the process requires a fusion plant that doesn’t exist yet.
This matters beyond alchemy. Many fields rely on isotopes from aging reactors: cancer drugs, reactor fuel, fusion plants, and deep-space batteries all depend on scarce isotopes. Marathon Fusion’s process could unlock a world of abundant isotopes, opening up many markets. For engineers, the practical takeaway is to watch for advances in affordable isotope production - if the fusion plant problem gets solved, the supply chain for medical isotopes and space power changes overnight.
Building Always-On Agents: The Primitives That Make Them Work
Always-on agents run indefinitely, managing context across sessions and taking action without direct supervision, per Claude’s engineering blog. They are built on five core primitives: an event inbox, timers, memory, permissions, and an outbox. They wake up periodically for runs rather than being truly always on.
Recent model advancements have made them more practical, with Opus 5.5 costing 78% less per solved task than Opus 5 on long-horizon coding benchmarks. The architectural pattern is worth stealing even if you’re not building on Claude: stateful agents with explicit permission boundaries and an outbox for side effects are the only way to run autonomous systems without them becoming a liability. The cost drop is the real story - 78% cheaper per task is the difference between a demo and a production service.
The Avalanche That Hit Math Is Coming for AI Development
AI’s capabilities in mathematics exploded quickly, and that ability to solve well-scoped problems is going to be an even bigger deal for machine learning research than for mathematics, per Dwarkesh Patel. METR found that Anthropic’s progress on well-scoped tasks is 2x what it was a year ago, largely due to AI use in R&D. OpenAI researchers’ heaviest 10% of AI users went from $1,000 to over $7,000 per day in coding agents by end of August, and Anthropic reported Claude leads over 26% of its AI R&D work, up from under 1% in February.
The author now believes AIs as good as the best humans at AI research are only months to a couple of years away, potentially leading to superintelligence by 2030. This is the recursive self-improvement loop made concrete: AI tools are directly speeding up AI development, which produces better AI tools. For engineers, the practical signal is that the cost of coding agents is dropping while their capability is compounding - the bottleneck is no longer the model, it’s your ability to integrate it into your workflow.
Trend Extrapolation Gets Us to AGI in 1-4 Years
Source: benjamintodd.substack.com ↗
Trend extrapolation suggests broadly human-level “digital worker” AGI could arrive in 2-4 years, with superhuman AI soon after, based on AI revenue growth, per Benjamin Todd. OpenAI and Anthropic are earning over $115 billion annualized as of September 2026, with revenue tripling every year for four years, and growth rates accelerating from 3x in 2024 to 6x in 2026.
If that continues, revenue could reach $10 trillion by March 2029, which would likely mean AI can do most remote jobs. The burden of proof is shifting to showing progress will stop. This is the most aggressive timeline in the newsletter, and it’s worth taking seriously even if you think the extrapolation is naive - the revenue numbers are real, and they’re the strongest signal we have that the capability curve isn’t flattening.
Tesla Opens Cybercab Pre-Orders: $29,990 Plus $299 a Month
Tesla opened pre-orders for its autonomous Cybercab “for personal use” to select customers, with a $29,990 price tag and a $299 per month operating fee, per TechCrunch. The vehicle is geofenced to certain Texas cities, with an estimated delivery date of December 2026. Elon Musk said the fee is “too high,” suggesting it should be ~$115/month.
Tesla also announced it will cease Cybertruck production next year to build more Cybercabs. The geofencing and monthly fee are the real constraints - this is a robotaxi with training wheels, not a personal vehicle. For anyone running fleets, the pricing structure is worth studying: a $299 monthly operating fee on top of purchase price is a new model for monetizing autonomy, and it’s clearly designed to test what the market will bear.
AI and Hiring: Bad for Employees, Bad for Employers
Source: johnhempton.substack.com ↗
The author, from financial firm Bronte, argues AI is bad for entry-level employees and also disrupts employers’ ability to find good staff, per John Hempton’s Substack. Bronte interviews all personalized CVs but quickly, often in five minutes, and categorizes longer interviews into genuine interest, learning, or “good-looking contra” - where they can profit from misconceptions.
This is a first-person account of hiring in an AI-affected market. The takeaway for engineers is that AI is flooding processes that require input from the public, causing many to break. If you’re hiring, you need to assume a percentage of your applicants are AI-generated; if you’re applying, you need to assume the screener is AI. The middle is getting squeezed.
China Fires Researchers After Child Dies in Gene-Editing Trial
Chinese authorities fired two researchers and disciplined several others after an investigation found safety and ethical flaws in an experimental gene-editing study that led to a child’s death, per Bloomberg. The two key researchers at Shanghai Jiao Tong University School of Medicine were barred for five years from government science and technology projects.
Several others received penalties including removal from posts, demotions, and disciplinary warnings. The researchers failed to flag warning signals from a non-human-primate study and didn’t disclose the child’s death in the published research paper. For anyone in regulated industries, this is a reminder that safety failures have career consequences, and that oversight is tightening globally.
Apple Launches AI Health Study App
Apple launched a new app, the Apple Health AI Study, to understand how people interact with conversational health AI tools, examining comprehension, response quality, trust, and health actions, per CNET. The study is open to 200,000 US participants aged 21+, requires an iPhone 15 Pro or later running iOS 26, and partners with OpenEvidence, which uses medical sources like NEJM and JAMA.
Data will not be used to train AI models but may inform feature development. This is Apple’s careful play: study before shipping, keep data on-device, and partner with credentialed medical sources. For engineers, the requirement of an iPhone 15 Pro or later running iOS 26 tells you where Apple thinks on-device AI processing needs to be - and that’s a hardware floor you should plan around.
You May Also Like
Nvidia's Huang Out, Musk Merger Looms, and Rust Finally Gets `!`
Jensen Huang is out as Nvidia CEO after two extramarital affairs were revealed, with CFO Colette Kress stepping in as interim chief. SpaceX and Tesla are …
Nvidia's Board Hunts for Huang's Replacement as Agents Take Over AWS
Nvidia's board has hired an outside firm to identify CEO candidates as it weighs ousting Jensen Huang following reports of extramarital affairs, including one …
Nvidia's CEO is gone, $1T vanished, and Apple wants Siri to actually do things
Nvidia CEO Jensen Huang resigned after admitting an affair with a woman US intelligence identified as a Chinese spy, wiping out over $1 trillion in market value …




