This Week in Tech 135

What we explore this week:

Naval Ravikant put the current moment simply: when software was expensive, thin horizontal tools extracted rents from every industry. Now that software is cheap, value collapses into vertically integrated players who own the full stack. Think Toast for restaurants, Veeva for pharma, or whatever AI-native startups are quietly building end-to-end in your sector right now. The middleware layer, the companies selling generic tools to everyone, is the clearest casualty of the current cost collapse.

Andrew Ng is making the same argument one level deeper. In a 15-minute dev conference talk, he claimed 100% of his own tasks are now handled by AI agents, and declared that within 3 to 6 months, building graph-based multi-agent architectures will be the standard workflow.

An Anthropic engineer made the design philosophy explicit: you are not supposed to monitor the model. Put it in a graph, and it catches its own errors while running a dozen tasks simultaneously. The company's published agent-building guidance reflects the same logic: parallel graphs, mutual error-checking, self-recovery. This is the production architecture pattern that turns AI from an impressive demo into something you can actually trust with real workflows. The teams that internalize this first will have a structural advantage over everyone still babysitting a single chatbot.

Claude developers confirmed Opus 5 is live in Claude Code and the broader Claude Platform today. Anthropic is putting its most capable reasoning directly into the environment where engineers actually work, not holding it back for API access. If the pattern from previous Opus releases holds, expect meaningfully stronger complex reasoning, longer context handling, and coding performance that puts real pressure on GPT-4o. Claude Code's documentation already reflects the update.

Most AI tools ask you to trade your data for intelligence. Norton Neo doesn't. It's the first safe AI-native browser built by Norton, and it gives you powerful built-in AI without handing your privacy over to get it. Search, summarize, and write with AI built directly into your browser. Your data stays yours. Your context stays private.

Built-in VPN, anti-fingerprinting, and ad blocking come standard. No add-ons. No setup. No compromises.

Fast. Safe. Intelligent. That's Neo.

Download Norton Neo

OpenAI made the keyboard optional. ChatGPT Voice is now in the desktop app, letting you speak to coordinate multiple AI agents running in ChatGPT Work or Codex simultaneously. Powered by GPT-Live, the app listens, responds, and executes in real time. Voice-controlled multi-agent coordination is no longer a research concept. It shipped today, globally.

A math PhD student used GPT-5 to knock out six open Erdős problems in five days. These are problems that had stumped professional mathematicians for decades. Polymarket flagged it, and Quanta has been tracking this trajectory for over a year. If the results hold up to peer review, this is AI doing original mathematical research, not assisting it. The field of formal mathematics is about to look very different.

Altman said it plainly: AI has entered its singularity moment. When the CEO from one of the world's most prominent AI lab uses that specific word in public, it is worth sitting with. The term has a precise history and a long record of being declared prematurely. Whether Altman is right or performing confidence for strategic reasons, the fact that singularity language is now mainstream from the people actually building these systems changes how the rest of the world is going to respond. Regulators, investors, and the public will all recalibrate around that word.

Nikita Bier pointed out the framing well: photos and video were how the last generation expressed themselves online. Software may be next. Grok's new app builder lets users generate playable apps and games directly inside the X timeline. TechCrunch confirmed the feature. If AI-generated micro-apps become a viral content format, the next Wordle comes from a prompt, not a dev studio.

Ilya Sutskever co-built the intellectual foundation of modern AI at OpenAI, then left quietly after the 2024 board crisis and founded Safe Superintelligence Inc. with a single focus: build superintelligent AI safely, with no product distractions. NVIDIA just wrote SSI a $5 billion check, per Reuters, at a $32 billion valuation. That is not a bet on a product. That is a bet on a person and a research direction.

The research direction is becoming clearer. Sutskever's public argument: humans are not born knowing everything, and that is the point. A system that arrives finished, fully pretrained, skips the trial-and-error loop that makes intelligence adaptive. Real intelligence comes from ongoing learning, not from a large snapshot of the world taken at training time.

X Money launched in the US, confirmed by The Verge, and the feature set is more complete than most expected: link bank accounts, move money, spend with an X debit card, earn interest, and eventually receive salary deposits, all inside one app. Visa confirmed it is powering real-time funding and payouts through Visa Direct alongside a co-branded X Visa card. The fuller picture is that X is not building a Venmo clone. It is building the financial operating system that makes everything else on the platform stickier: payments, savings, spending, all in one place that already has your attention daily.

Apple TV+ confirmed a January 22 premiere for Neuromancer, with the first-look poster shared by director JD Dillard. Dillard's own post made clear this is his project in full. William Gibson's 1984 novel invented the word cyberspace, named the matrix before the Wachowskis borrowed it, and seeded the hacker aesthetic that runs through every piece of tech culture you recognize today.

It also predicted corporate surveillance, AI consciousness, and the collapse of the line between digital and physical identity with uncomfortable precision. Wired's 40th anniversary piece lays out just how much of the present Gibson described from a 1980s typewriter.

Polymarket flagged the official release of Augmental's MouthPad, a retainer-style device that lets users control computers and AI agents via tongue movements. The primary use cases are accessibility and hands-free computing, though the "vibe coder" framing tells you where the cultural conversation around human-computer interfaces is heading. IEEE Spectrum covered the hardware in detail. The idea that a retainer could replace a mouse is strange until you remember that every input paradigm looked strange before it was normal.

Sneaker News broke the Nike x Hyperice recovery slides: active massage and vibration therapy built directly into footwear. Hyperice has been pushing recovery tech into mainstream sports for years. Putting it in a slide you wear around the house after a workout is the most accessible form factor yet. If the product performs, it reframes what a rest day looks like, and opens a new product category at the intersection of footwear and wellness hardware.

Palmer Luckey livestreamed the M64 launch, Anduril's autonomous counter-drone platform designed to track and neutralize drone swarms at scale. Doing a live product stream for defense hardware is a deliberate Silicon Valley move: Anduril is normalizing the idea that weapons systems get hype drops like GPUs. The counter-UAS market is growing fast given the role drone warfare has played in recent conflicts. The M64 is Anduril's most public statement yet that it intends to own that category.

Two ideas kept surfacing in different forms across this week's stories, and they are really the same idea stated from different angles. Naval says cheap software kills the middleware layer and hands power to whoever builds the deepest end-to-end experience. Ng says the architecture you choose matters more than the model you pick. Both are making the same underlying argument: the abstraction layer that used to be the product is disappearing, and what replaces it is systems thinking. The builders who understand how to wire things together, how agents check each other, how feedback loops substitute for raw intelligence, are the ones who will build what comes next. Prompt engineers are the new data-entry clerks: useful for a moment, then absorbed into the workflow.

Sutskever's bet at SSI adds a third angle to the same picture. If models that stop learning after training are fundamentally limited, and SSI is right that continual learning is solvable, then the current generation of AI products are early drafts. Expensive, impressive early drafts, but drafts. A system that keeps learning from experience, that adapts the way a person does, is a qualitatively different thing from a very large lookup table. NVIDIA writing a $5 billion check to the person most credibly working on that problem suggests the industry's smart money agrees. The race is no longer just about who has the biggest model. It is about who figures out what comes after pretraining.

Neuromancer arriving on January 22 is not a coincidence worth ignoring. Gibson wrote that book in 1984 on a manual typewriter, imagining a world where corporations controlled AI, where the line between human consciousness and digital identity was blurry and contested, and where the people who understood the architecture of the network held real power. We are having all of those arguments right now, in public, at scale. X Money, SSI, agent graphs, voice-controlled computers: Gibson mapped the territory before anyone had built the roads. The show is coming at the exact moment his themes have moved from science fiction to the front page. That is either great timing or a reminder that the best sci-fi was never really about the future.

Markets don't wait for Monday. News breaks on a Saturday morning, and most traders can do nothing but watch.

Not on Liquid. Trade domestic and international equities, commodities, forex, crypto, and prediction markets — all from one account, 24 hours a day, 365 days a year. Liquid gives you access to any market, from anywhere, anytime. To us, access is arbitrage.

Getting started takes under 10 minutes: log in with Google, deposit with Apple Pay or a bank transfer, and trade from your phone or desktop — wherever you are in the world.

While everyone else is refreshing headlines and waiting for the open, you're already positioned. That's the difference between reacting to markets and actually trading them.

Trade 24/7 on Liquid Today

Making AI actually work day to day is becoming its own job. Hear from three people doing it: Simone Santiago Broad (Yoco), Yelva Espinoza (Zumba Fitness), and Fin's Dave Lynch. They share what the role really looks like, how it came to exist, the skills worth hiring for, and the challenges they're tackling right now. Watch the full conversation on demand.

Watch now