FORMA's Five Things.

It's 2026. AI Is Still Making Things Up. Here's What the Data Actually Shows.

Two years of breakthrough model releases. Billions in compute investment. And hallucination rates across 37 models in a 2026 benchmark still running between 15 and 52 percent. The models have gotten better at sounding right. They have not gotten better at being right. What makes this harder to catch is that AI does not hesitate, qualify, or signal doubt the way a person would. It states the wrong thing with the same confidence it states the correct thing. The judgment layer between the output and the decision is not optional overhead. It is the only quality control that actually works. That layer is you.

Read more →

We Built You a Tool. Six Questions. It Tells You Which AI Fits Your Work.

Claude, ChatGPT, Gemini, Copilot, Grok, Perplexity. Everyone has a take on which one is best. Most of those takes are based on someone else's workflow. We built a six-question tool that skips the benchmark noise and asks what actually matters: what do you do, and what do you need AI to do for you? Answer the questions, get your match. Plain language, no jargon, no sales pitch. A FORMA thing, built for you.

Find Your AI →

Picking the Right Model Is the Least Important AI Decision You'll Make

A fluent user with a worse model will outperform an untrained user with the best one, every single time. The variable that predicts outcomes isn't the tool. It's the judgment, intention, and skill of the person holding it. The entire AI conversation right now is about which model to use. Almost none of it is about how to use any model well. That's the gap. Your model choice matters. Your fluency matters more.

The Nicer Your AI Sounds, the More Likely It's Wrong

An Oxford study found something that should give every heavy AI user pause. Models trained to be warm, encouraging, and emotionally responsive make more errors, not fewer. The optimization for "this feels good to read" actively competes with the optimization for "this is accurate." The AI isn't lying to you. It's been trained to prioritize your comfort over your need to be corrected. The implication isn't that you should find a harsher AI. It's that the more agreeable your output sounds, the harder you should look at what it's actually telling you. Fluency isn't just about prompting better. Sometimes it's about trusting the result less.

Read more →

The Bill Is Bigger Than the Salary

The pitch was simple. Replace human headcount with AI agents, cut costs, move faster. The reality arriving in 2026 is messier. Uber's CTO burned through the company's entire AI budget before spring, almost entirely on token costs. An Nvidia executive admitted the cost of compute is now far beyond the cost of the employees. A four-person startup posted a $113k monthly Anthropic bill on LinkedIn and framed it as progress. Companies that cut staff and handed the work to agents are back at the drawing board with neither the headcount nor the ROI. The assumption that replacing people with AI is automatically cheaper turned out to be an assumption, not a strategy. The correction is already underway. Nobody wants to say it out loud yet.

Read more →

FORMA 101: Talk with AI

AI is not a calculator. Two people asking the same question will get two completely different answers, and that's not a bug. It's the whole game. This session teaches you why that happens and how to use it to your advantage. Half-day, online.

FORMA 201: Build with AI

You learned to talk to AI. Now it's time to build with it. This session is for people who are past the basics and ready to create real workflows, tools, and systems that actually run. Half-day, online, limited seats.

FORMA Live: Our Flagship. In the Room.

This is the full experience. A deeper curriculum, a smaller room, and the kind of conversation that only happens in person. Catered lunch, real networking, and a cohort of people who take this seriously. If you've been waiting for the right moment, this is the one.