FORMA's Five Things.

FORMA Programs

If you could sit down with AI for two hours
and walk away knowing how to actually
build something, what would you build?

That is the question we are designing the next Build With AI session around. Not what we think you should learn. What you actually want to get your hands on. Claude Code. Cowork. AI Dashboards. Deep Research. You tell us, and we will build the session around it.

The US Government Just Pulled
Anthropic's AI Models Offline.
Can They Do That?!

This week's newsletter has a theme, and the theme is this: the relationship between AI and the people in charge of the rules just got very, very real.

Last Friday evening, the US government told Anthropic to shut off access to its two newest AI models, Fable 5 and Mythos 5. The order said no foreign nationals could use them. Not overseas. Not in the US. Not even if they work at Anthropic. The problem was, Anthropic had no way to block just foreign users. So to comply, they pulled both models for everyone. Days after launch, they went dark.

The stated reason was national security. Another company claimed it found a way to bypass the models' safety features, and the government moved fast. Anthropic says it was given almost no details and disputes how serious the issue was. But the decision was already made.

Here is what makes it stranger. This is the same company the Pentagon blacklisted months ago for refusing to let the military use its AI for autonomous weapons. So now Anthropic is simultaneously too dangerous for the government to use and too dangerous for anyone outside the US to touch.

The signal is clear. AI models are no longer just apps you subscribe to. The government is treating them like weapons technology. And the companies that build them may not get the final say on who uses them.

Read more →

Congress Just Dropped 269 Pages
of AI Rules. And They Mean It.

While the Fable story grabbed headlines, something quieter landed on June 4 that may matter more in the long run.

Representatives Jay Obernolte and Lori Trahan released a discussion draft of the Great American Artificial Intelligence Act. It is 269 pages. It is bipartisan. And it is the first serious attempt at a comprehensive federal AI framework in the United States.

The bill targets frontier models and introduces safety reporting requirements, whistleblower protections, workforce impact tracking, and new rules around AI-driven layoffs. It would also preempt state AI laws for three years, which has already sparked a fight.

Colorado's AI Act takes effect June 30. Companies that delayed compliance waiting for Washington to step in are now 22 days from a real enforcement date with no federal cover. The rules are coming. The only question left is who writes them first.

Read more →

There Is a War Coming.
An AI Price War Is on the Horizon.

Google just dropped its entry-level AI subscription to $4.99 a month. OpenAI is reportedly preparing to slash token prices. Anthropic is expected to follow. After two years of competing on who had the smartest model, the AI industry is about to compete on who can charge the least.

If this feels familiar, it should. In the early 2000s, cell phone carriers went through the exact same cycle. First they competed on coverage. Then features. Then it became a pure price war. Unlimited minutes. Family plans. Free nights and weekends. The ones who survived figured out how to make the math work at scale. The ones who did not got bought or disappeared.

That is where AI is heading. For most business tasks, these models perform similarly enough that the smartest one does not matter as much as the cheapest one. CIOs are treating them like interchangeable utilities. Loyalty to a brand evaporates when you can just switch to whatever costs less that day.

The catch is that AI is wildly expensive to run. OpenAI and Anthropic are both losing billions a year. Google can absorb it because ads print money. The startups cannot. And both are heading to IPO. If you are a buyer, the price war is great news. If you are a seller, the math just got uncomfortable.

Read more →

Bots Just Passed Humans on the Internet.
And Bots Don't Click Ads.

Cloudflare's CEO confirmed it this month. For the first time ever, automated bot traffic has surpassed human traffic online. The split: 57% bots, 43% humans. He thought this would not happen until 2027. It happened now.

The real story is not the volume. It is what it breaks. The entire economic model of the internet is built on one assumption: a human is on the other side of the screen. That human sees a banner ad. Scrolls past a sponsored post. Clicks a search result. Every impression, every performance campaign depends on a real person actually being there.

AI agents change that completely. These are not the old fraud bots programmed to click on things and inflate numbers. AI agents browse the internet on your behalf, which means you never load the page yourself. You never see the ad. The human attention that the entire ad economy is selling just left the building.

You cannot solve this with better fraud detection, because nothing fraudulent is happening. The bot is doing exactly what it was asked to do. Performance marketing has spent years optimizing for clicks, impressions, and conversions. It may be about to find out how many of those metrics required a human to mean anything at all.

Read more →

You Are Being Charged by a Unit
You Have Never Heard Of

If you have used ChatGPT, Claude, or Gemini, you have been charged by the token. You just might not know it yet.

A token is the basic unit AI companies use to measure how much of their product you are consuming. It is not a word, exactly. It is more like a fragment. The word "hamburger" is two tokens. The word "AI" is one. A typical paragraph runs about 100 tokens. Every message you send costs tokens. Every response you receive costs more. And the output tokens, the ones the model generates, are almost always more expensive than the input tokens, the ones you type.

Here is why this matters right now. The biggest companies in AI are in the middle of a price war, and most people using these tools have no idea what a token is, how their usage is measured, or why one conversation costs more than another.

Here is something else most people do not realize: you do not need the most powerful model for every task. Need to clean up an email or summarize meeting notes? A smaller, cheaper model handles that just fine. Save the frontier models for the heavy lifting, the complex analysis, the work that actually demands it. That one habit alone can cut your token costs dramatically without changing the quality of your output.

FORMA's take: before you compare prices, learn what you are paying for. The token is the unit. Matching the right model to the right task is the strategy. Start there.

FORMA Tools

Not sure which AI model is actually right for your work?

Six questions. A personalized model recommendation. No benchmark tables, no guesswork. Just a clear answer based on how you actually work.