Find out why 100K+ engineers read The Code twice a week

Staying behind on tech trends can be a career killer.

But let’s face it, no one has hours to spare every week trying to stay updated.

That’s why over 100,000 engineers at companies like Google, Meta, and Apple read The Code twice a week.

Here’s why it works:

• No fluff, just signal – Learn the most important tech news delivered in just two short emails.

• Supercharge your skills – Get access to top research papers and resources that give you an edge in the industry.

• See the future first – Discover what’s next before it hits the mainstream, so you can lead, not follow.

Join 100,000+ engineers who read The Code to stay ahead of the curve.

Hi there,

It’s Tuesday, January 20, 2026. I hope your year is off to a strong start.

If you’ve been feeling like the news cycle is moving faster than ever this month, you’re not alone. 

We have a lot to cover today, so grab your coffee, and let’s dive into what is actually happening.

Today:

• Anthropic’s Economic Reality Check: "Deskilling" Shock and the Rise of Agents

• Google’s Gemini Sales Explode as AI Starts Handling Commerce

• OpenAI Report: Revenue Now Scales Perfectly With Intelligence

• Ex-OpenAI Policy Chief Launches Independent Safety Auditor 

• Meta Hires Dina Powell McCormick to Spearhead AI Infrastructure 

• Chinese Unicorn Moonshot AI Jumps to $4.8 Billion Valuation

"Deskilling" Shock is Coming | Anthropic Economic Report

Anthropic’s latest AI economic report paints a more grounded picture of automation's rise. Agentic AI is real—tools like Claude Code and Co-Work are already doing multi-step tasks, from coding to planning. But full automation isn’t arriving overnight. Coding leads the way, but other sectors lag. 

Instead of job destruction, many roles are shifting: some get “deskilled” as AI handles complex work, while others are “upskilled,” freeing humans to focus on higher-value tasks. Productivity gains are real but smaller than hyped—especially without human oversight. The future of work looks more like managing smart assistants than being replaced by them. Slower… but still seismic.

WATCH THE VIDEO ON YOUTUBE

Google’s Gemini business is quietly getting loud

Google’s enterprise Gemini push looks like it’s turning into real usage, not just hype. The clearest signal: API demand. A report citing internal data says Gemini API requests jumped from ~35B in March to ~85B by August (more than doubling in five months), with the climb starting after Gemini 2.5 and continuing with Gemini 3.

A few extra details that make this more interesting than a raw “big number” headline:

• Google says Gemini Enterprise has 8 million subscribers across ~1,500 companies, plus another ~1M sign-ups online.

• Feedback is split: some users love the “connect to company data” angle, but others say it’s solid for basic Q&A and weaker on specialized tasks and custom app building.

• This sits on top of Google’s earlier “Gemini Enterprise” positioning as a business platform you can chat with across company data/docs/apps.

Why it matters: if these numbers hold up, it’s the clearest sign yet that Google isn’t just “catching up on models,” it’s converting distribution + enterprise packaging into sustained demand. That’s the difference between a demo race and a revenue race.

OpenAI just laid out the most explicit “AI scales like a utility” business narrative yet

Sarah Friar’s new OpenAI post is basically a blueprint: the company says its business model is designed to scale with delivered value — subscriptions, usage-based APIs, and now commerce + advertising inside ChatGPT (with the emphasis that it needs to be labeled and genuinely useful).

Two lines in there are doing a lot of work:

• OpenAI says revenue tracks available compute and shares a simple arc: compute ~0.2 GW (2023) → 0.6 GW (2024) → ~1.9 GW (2025), while revenue grew from $2B ARR (2023) → $6B (2024) → $20B+ (2025).

• The 2026 focus is blunt: “practical adoption” — closing the gap between what AI can do and what people actually use day-to-day.

Under the hood, the strategic shift is “compute as a managed portfolio”: train frontier where performance matters, serve high-volume workloads on cheaper infrastructure where efficiency matters, and keep contracts flexible across providers/hardware.

Why it matters: this is OpenAI telling the world, “We’re not pricing intelligence like software seats forever — we’re pricing it like outcomes and throughput.” It’s also a not-so-subtle message to competitors: distribution is nice, but compute certainty is the moat.

A former OpenAI policy leader is calling out the “trust gap” — and trying to build an audit industry

Miles Brundage (former OpenAI policy research lead) has launched AVERI, a nonprofit focused on making third-party auditing of frontier AI “effective and universal,” and pushing back on the industry norm of companies effectively grading their own homework.

AVERI’s definition is important: auditing means rigorous third-party verification of safety/security claims, evaluated against standards, with deep, secure access to non-public information.

They also propose AI Assurance Levels (AAL-1 to AAL-4) so audits aren’t just vibes — they’re legible and comparable.
And they argue audits should cover four risk buckets:

1. intentional misuse (ex: cyberattacks)

2. unintended system behavior (ex: harmful errors)

3. information security (ex: model/data theft)

4. emergent social harms (ex: addiction / facilitation of self-harm)

Why it matters: the market is starting to demand “trust you can prove,” not “trust our blog post.” If enterprise buyers, insurers, or regulators start requiring credible audits, that becomes a new kind of gating function for frontier labs.

🧠RESEARCH

The Poisoned Apple Effect: Strategic Manipulation of Mediated Markets via Technology Expansion of AI Agents

Researchers identified a sneaky tactic called the "Poisoned Apple" effect where companies release AI tools they don't intend to use, just to force regulators to change market rules in their favor. This manipulation shows that current laws are too slow and must adapt quickly to how AI agents actually compete.

Unlocking Implicit Experience: Synthesizing Tool-Use Trajectories from Text

Teaching AI to use software tools usually requires expensive, hand-made data. A new method called GEM solves this by extracting problem-solving steps hidden in regular text documents to create training lessons. This approach allows AI to learn complex, multi-step tasks from existing internet content without needing human supervision.

ShapeR: Robust Conditional 3D Shape Generation from Casual Captures

Creating 3D digital objects usually requires perfect studio photos. A new system named ShapeR can build detailed 3D models from shaky, messy videos taken on regular phones. It uses clever software tricks to ignore background clutter and missing details, making it much easier to turn real-world items into virtual ones.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

Casper AI – Chrome Summarizer for Faster Reading & Research - Chrome extension that uses AI to quickly explain and summarize web pages and PDFs, aimed at professionals who need to digest information and share takeaways efficiently.

CassetteAI – Fast Text‑to‑Music Generation for Royalty‑Free Tracks - web and iOS tool that turns text prompts into full music tracks and sound elements.

Castmagic – Turn One Recording into 100 Content Assets - AI content workspace that converts long‑form audio or video into ready‑to‑publish assets—transcripts

📲SOCIAL MEDIA

🗞️MORE NEWS

Meta hires political heavyweight for AI push Meta has appointed Dina Powell McCormick, a former Trump adviser and Goldman Sachs executive, as its new president and vice chairman. She will oversee the company's massive plan to build the physical systems needed for artificial intelligence, such as data centers and power plants. This hiring suggests Meta is prioritizing strong political and financial connections to secure the resources it needs to lead the industry.

Chinese AI startup hits $4.8 billion value Moonshot AI, a startup backed by Chinese tech giants Alibaba and Tencent, has seen its value jump to $4.8 billion in a recent fundraising round. While its local competitors have rushed to sell shares to the public to raise money, Moonshot is choosing to stay private as its price continues to climb. This rapid growth highlights the intense demand for domestic AI tools in China, where American services like ChatGPT are not available.

New data to fix awkward AI translations Hugging Face has released a massive collection of translated text designed to help computers learn languages more naturally. The project aims to eliminate "translationese"—the stiff, unnatural phrasing that often results when AI translates words literally rather than capturing the intended meaning. This effort is particularly focused on improving quality for languages that currently lack enough data for effective AI training.

Microsoft identifies jobs most "exposed" to AI A new report from Microsoft Research lists 40 professions most likely to be affected by artificial intelligence, specifically targeting roles that involve writing, scheduling, and processing information. The researchers clarify that being "exposed" does not mean these jobs will disappear, but rather that the daily tasks will change drastically as AI handles the routine work. The study suggests that workers in these fields should focus on learning to work alongside these tools to stay relevant.

New leaders drive Sequoia’s Anthropic deal Sequoia Capital’s new leadership team has approved a first-time investment in Anthropic as part of a massive $25 billion fundraising effort. This marks a sharp reversal in strategy, as the firm’s previous boss had refused to back the company due to loyalty to other AI investments. The deal is being led by major global investors and is expected to double Anthropic’s value in just a few months.

Crisis at Mira Murati’s AI startup Thinking Machines, the AI company launched by former OpenAI executive Mira Murati, is in turmoil after she fired a co-founder and several top researchers quit to join a competitor. These sudden departures have shaken the confidence of investors right as the company attempts to raise money at a $50 billion valuation. The internal chaos raises serious questions about the startup's stability and leadership just one year after it was founded.