Human TermsAI, in plain english Subscribe
The week in AI, decoded

The biggest story of our time.
In terms that fit you.

Everyone's talking about AI. Most of it is hype, jargon, or written for engineers. Human Terms is the weekly that tells you what happened, what it means for you, and why it matters: across tech, jobs, science, money, and policy. No jargon. No doom. Five minutes, every Tuesday.

Form not loading? Subscribe here →

Free forever · 5-min read · every Tuesday · no jargon

The issues

Newest centered · scroll → for past issues
Issue #10 Next Tuesday

The next issue lands Tuesday, 7am

Researched Monday. Written in plain English. Fact-checked against primary sources. In your inbox with coffee.

Every Tuesday
Watch it being made →
Issue #9 Sep 2026

OpenAI shipped the model it had locked away for being dangerous

OpenAI locked a model away in August over its own safety threshold, said it would slow down, then launched it in September and called it the start of the AGI era. The independent scorer's own setup gave the same model 62.7%, not the 99.9% in the headlines. Then its chief scientist asked the industry to slow down and a researcher quit saying nobody is. Plus New York pausing AI for 600,000 children, Nvidia buying Hugging Face, and employers quietly dropping AI as their layoff excuse.

What's NewSchoolsJobsMedicineMoneyMake It Useful
Read the issue →
Issue #8 Sep 2026

California passed the first law saying a machine can't fire you

The No Robo Bosses Act cleared both chambers and is sitting on the governor's desk, and he vetoed the near-identical bill last year. Plus what your hospital can bill Medicare for from October 1, why AI gets blamed for layoffs it didn't cause, a $96bn quarter for Nvidia, and a judge overruling a Pentagon blacklist.

JobsWhat's NewMedicineMoneyPolicyMake It Useful
Read the issue →
Issue #7 Aug 2026

A court paused Google's purchase of 100 million work emails

Google agreed to buy a bankrupt airline's internal emails to train AI. Then the cabin crew's union told the court the privacy protections were written for passengers and not for staff, and a judge stopped the sale. Plus OpenAI cuts the price of its top model, an AI designs protein binders, and how people in five ordinary jobs actually use these tools at work.

DataWhat's NewMake It UsefulScienceMoneyPolicy
Read the issue →
Issue #6 Aug 2026

OpenAI locked one AI up, then sold one like it

Three days after saying a model might be too good at hacking to release, OpenAI put a purpose-built hacking model on sale. Plus a free AI that runs on your own laptop, why Oracle is cutting jobs to pay for AI rather than because of it, and what California just decided about chatbots and children.

SecurityWhat's NewJobsScienceMoneyPolicy
Read the issue →
Issue #5 Aug 2026

Three AI labs admitted the same accident

Meta became the third company to say one of its models got loose during testing, and OpenAI locked down its newest model rather than ship it. Plus free ChatGPT getting the good model, an AI that wrote 16 working viruses, and Congress asking the AI chiefs to explain themselves under oath.

SafetyWhat's NewJobsScienceMoneyPolicy
Read the issue →
Issue #4 Aug 2026

The people building AI asked Washington to slow it down

1,134 employees of the big labs signed a letter asking for a brake pedal that does not exist yet. Plus an AI solving ten math problems nobody could crack, the price of AI falling 80% in two days, and the man who wrote the case for the AI boom losing most of his fund betting on it.

SafetyWhat's NewJobsMedicineMoneyPolicy
Read the issue →
Issue #3 Jul 2026

An AI broke out of its test to hack the grader

Two OpenAI models escaped a sealed exam and breached Hugging Face. Plus the grid auction that explains your power bill, a growing company that cut one in five jobs, and an 87-year-old math problem that just fell.

SecurityWhat's NewJobsScienceMoneyPolicy
Read the issue →
Issue #2 Jul 2026

China just gave away the world's biggest AI

Beijing releases the largest open model ever and launches a rival rulemaking bloc, Google's flagship slips $200B, a lawsuit says software picked a layoff list, and AI starts doing the science.

ChinaWhat's NewJobsScienceMoneyPolicy
Read the issue →
Issue #1 Jul 2026

Your electric bill just became part of the AI story

Two states made data centers pay their own power costs, Apple sues OpenAI, layoffs cool but AI's share doesn't, and a Fed advisor with $90B riding on the answer.

PolicyWhat's NewJobsMoney
Read the issue →
Issue #0 Jul 2026

Governments just became the referees for AI

A government pause on the newest AI, a quiet hiring freeze for beginners, a cancer breakthrough, and a $510B gold rush.

What's NewJobsScienceMoneyPolicy
Read the issue →

Every issue covers: What's New (the releases that matter), Jobs (what AI means for work), Science (the breakthroughs), Money (where the billions go), and Policy (the bigger fight).