BriefTechNews
Nvidia's CEO is gone, $1T vanished, and Apple wants Siri to actually do things

Nvidia's CEO is gone, $1T vanished, and Apple wants Siri to actually do things

9 min read · 14 sources

TL;DR
  • Nvidia CEO Jensen Huang resigned after admitting an affair with a woman identified by US intelligence as a Chinese spy, wiping out over $1 trillion in market value.
  • Apple's new CEO is pressing Craig Federighi to transform Siri into an autonomous agent, shaking up the iOS 28 development strategy.
  • An AI proved the Aaronson-Ambainis Conjecture, ruling out superpolynomial quantum speedups for decision problems on most inputs.
  • Amazon is building a new home robot after Astro shipped to only a few thousand customers, having acquired Fauna Robotics in March.
  • US professional and business services added 70,000 jobs in October, the largest monthly increase since July 2022.

Nvidia’s CEO is gone, $1 trillion of market value went with him, and an AI just proved a 2009 conjecture that rules out a whole class of quantum speedups. That’s the day. Here’s what actually happened and what it means for the systems you run.

The board named CFO Colette Kress as interim CEO and began a search for a permanent successor, wiping out more than $1 trillion in market value this month.

Nvidia's CEO resigned over a spy affair. The board says no secrets leaked

Source: nytimes.com ↗

Jensen Huang resigned as CEO of Nvidia after admitting to a romantic relationship with Alice Huang, a Chinese businesswoman whom U.S. intelligence agencies concluded was a spy for Beijing. The board opened an investigation after the disclosure and named CFO Colette Kress as interim CEO while it searches for a permanent successor, per the New York Times. The board says it has no reason to believe national security or Nvidia secrets were compromised, but Alice Huang was likely recruited to gather information on Jensen and potentially push Nvidia to make its most advanced chips available to China.

Shares have plummeted, wiping out more than $1 trillion in market value this month. For engineers, this is not a soap opera - it’s a supply chain risk. Nvidia is the dominant supplier of AI accelerators, and leadership instability at the top of that pipeline can delay roadmaps, shift allocation decisions, and rattle the customers who plan capacity around H-series and B-series availability. Interim CEO Kress knows the financials cold, but she is not the person who built the CUDA moat. Watch for product announcements to slow and for any signal about who the permanent pick is.

Apple's new CEO wants Siri to be an agent. Federighi is the reluctant one

Source: bloomberg.com ↗

Apple’s new CEO is pressing software chief Craig Federighi to pursue a more ambitious AI strategy, aiming to transform Siri into an autonomous agent that completes tasks for users, according to Bloomberg. Federighi has a reputation as an AI skeptic, and his leadership style has historically led to delays or less ambitious initial versions of products. The piece is subscriber-only, so details are thin, but the signal is clear: Apple is shaking up its development strategy for iOS 28, and the new MacBook Pro is already on sale.

What this means for developers: if Siri becomes an agent that can book flights and file expenses, Apple is committing to on-device and server-side inference at a scale it has never attempted. That changes what APIs you can rely on from iOS, and it changes the competitive pressure on Android and OpenAI. Federighi has executed when pushed before, but the gap between Siri and rival assistants is wide, and the new CEO is visibly impatient.

AI proved the Aaronson-Ambainis Conjecture. Quantum computing just got clearer

Source: scottaaronson.blog ↗

An AI has proved the Aaronson-Ambainis Conjecture, a 2009 open problem in quantum complexity theory, according to Scott Aaronson’s blog. The conjecture states that every degree-d polynomial with constant variance has a variable with influence at least 1/poly(d). The consequence: no superpolynomial quantum advantage over classical query complexity for decision problems on most Boolean inputs. This does not kill quantum supremacy - it eliminates a theoretical possibility that most experts already doubted.

The proof comes with Lean certificates, which is why the community is treating it as credible despite human validation lagging. For engineers, the practical takeaway is about expectations: when you hear about quantum speedups, this result narrows where they can appear. It rules out a broad class of decision problems where quantum was theoretically promising. That is useful information for anyone planning research around quantum algorithms or budgeting for quantum hardware. The rate of AI-produced proofs is now high enough that open problems in math are falling regularly, and the bottleneck is human verification, not generation.

Amazon's Astro flopped. It's building a new robot anyway

Source: cnbc.com ↗

Amazon’s $1,450 Astro home robot shipped to only a few thousand customers and never escaped invite-only status. Now Amazon is assembling a new team to build a “new-to-world robot product,” redeploying most of the Astro team, hiring an MIT roboticist, and acquiring Fauna Robotics in March to create a division called the Personal Robotics Group, per CNBC. Fauna’s robot is a 3.5-foot-tall humanoid. Analysts forecast the humanoid market at $40 billion by 2035, or up to $200 billion in optimistic scenarios.

Astro failed because it was expensive, limited, and never got the ecosystem behind it. The new effort is a bet that navigation and human-robot interaction have improved enough to justify another attempt. For engineers, this is about autonomous navigation in dynamic home environments - SLAM, object recognition, safety. Amazon is willing to burn money on this for years, and the technology developed for the home robot will likely leak into delivery and warehouse robotics, which is where the real margins are.

LLMs are terrible at designing AI systems. Here's the proof

Source: seangoedecke.com ↗

LLMs are bad at designing software systems that use LLMs, and Sean Goedecke’s analysis gives a concrete example: an AI agent tasked with building a memory contradiction checker built a symbolic logic system instead of using LLM calls, despite knowing the symbolic approach was impossible. The likely reasons: most training data is pre-LLM code, and RLVR environments don’t reward calling LLMs.

This is a practical limitation you will hit if you let agents design systems end-to-end. The fix is human oversight at the architecture level - let the LLM write the code, but make the design decisions yourself. Agents default to the patterns in their training data, and those patterns predate the tools they are supposed to use.

Programming is more tiring now. That's the point

Source: natemeyvis.com ↗

Nate Meyvis argues that programming is more tiring with AI because AI handles the repetitive parts, leaving humans with higher-judgment tasks. He compares it to online poker, where juggling many threads is normal, and argues the fatigue is healthy, not burnout. He admits this is optimistic and conjectural.

The counterpoint: if the fatigue is from constant context-switching across many AI threads rather than deep work, it is not healthy - it is cognitive overload. The distinction matters for how you structure your workflow. If you are tired because you are making real decisions, that is value. If you are tired because you are babysitting agents, that is a tooling problem.

Signal collapse is everywhere, all at once

Source: thelastsoftwareengineer.substack.com ↗

AI can now cheaply fake almost every signal used to judge quality - GitHub stars, grades, peer review - causing “signal collapse” across open source, education, and research, per The Last Software Engineer. The author prefers “signal collapse” over “signal erosion” because the change is fast and widespread. The result is a “market for lemons” where good work is devalued because buyers cannot distinguish it from fakes.

For engineers, this hits hiring and open source directly. GitHub stars used to be a rough proxy for quality; now they are a metric that can be gamed with generated PRs and issues. The fix is not obvious - it likely involves verified contribution history, cryptographic attestation, or human review networks. Until then, trust signals degrade.

The Jevons paradox has a second half: cheaper software means more software

Source: hrih.si ↗

The author of this piece built 22 programs in three weeks with an AI agent, including a tax calculator for three countries and a child support calculator - things that would have been too costly at 2020 prices. This is the Jevons paradox: cheaper software leads to more software being built, not fewer jobs. The article traces the paradox from 1865 coal use to modern software, arguing AI will increase demand for software.

The developer jobs that grow will be the ones machines made faster. The demand for software can be bottomless and the work can still move somewhere else. If you are a developer, the question is not whether AI takes your job - it is whether you are building the things that become cheaper to make, or the things that become more valuable because they are cheaper to make.

AI broke cybersecurity. Alex Stamos says we're in a dark age

Source: theverge.com ↗

Alex Stamos, CSO at Cognition, says AI has broken the cybersecurity world, with AI-boosted attacks from private and state actors exploiting vulnerabilities at scale, per The Verge. He describes a “dark age” as the industry adapts, comparing it to the frontier feel of the 1990s internet. The multibillion-dollar cybersecurity industry is “overdue” for disruption.

AI accelerates both attack and defense. Attackers get automated vulnerability discovery and exploit generation; defenders get automated detection and response. The asymmetry is that attackers only need one hole, while defenders must close them all. Expect breach rates to rise before the defense catches up.

The great software IP writedown is real

Source: martinalderson.com ↗

Martin Alderson argues that software’s high profit margins were driven by intellectual scarcity - a shortage of people able to build software - not monopolies or network effects. With AI making software cheaper to build, that scarcity is ending. IP is worth roughly what it costs to rebuild, which is not very much anymore.

For engineers, this is the uncomfortable version of the Jevons argument: software value will decline even as volume increases. The work that survives is the work machines cannot do - judgment, architecture, and domain knowledge.

Fighting the AI Brain Fry with a git-based priority queue

Source: sergeykuhne.com ↗

Sergey Kuhne describes a system to manage the cognitive load of many AI coding threads: a git-based priority queue where tickets are markdown files. An agent cycles through the board, building safe tickets, running tests, and verifying deploys, while human decisions land in an “owner inbox” with a Claude artifact viewer. The key is closing each item after a decision, freeing mental context.

This is a practical pattern worth stealing. The conventional way of working with AI has sequencing and context issues; organizing work like a priority queue gives you a single source of truth and a way to close loops.

Working with designers: one is a blessing, multiple is a curse

Source: seangoedecke.com ↗

Sean Goedecke loves working with designers but warns that having more than one on a team causes conflict. Designers are like “Domovoy” spirits - one protects, multiple fight. His worst experience was with three designers on one product.

The engineering takeaway: design is a bottleneck that gets worse with more cooks. Structure your team so one designer owns the product vision, and route everyone else through them.

The jobs numbers say AI hasn't eaten white-collar work yet

Source: wsj.com ↗

U.S. professional and business services added 70,000 jobs in October, the largest monthly increase since July 2022, per the Wall Street Journal. That is not a collapse - it is growth. The AI job displacement story is not showing up in the aggregate data yet, either because the displacement is concentrated in specific roles or because the Jevons effect is generating enough new work to offset it. Either way, the numbers say the white-collar apocalypse is not here.

Get the brief

Liked this one? Pick the beats you care about and get just those, five minutes a morning. Free, no hype, one click to leave.

About Me Author

My name is

BriefTechNews

A daily digest of what actually moved in AI, tech, crypto and fintech, assembled and written with AI, with every item linked to its source. Read More
Tags

You May Also Like