FORMA's Five Things.

Coming Up

May 20

Talk with AI

Virtual · 101

May 21

Build with AI

Virtual · 201

June 3

Talk with AI

In-Person · 101

El Segundo, CA

The AI Read the Resume. Loved It. Turns Out, It Wrote It Too.

Wanna hear something that is equal parts crazy and completely logical once you think about it? A new study out of the University of Maryland found that when a candidate uses the same AI model to write their resume as the company uses to screen it, they are up to 60% more likely to make the shortlist. Not because they are more qualified. Because the AI recognized its own syntax. The machines are out here hiring each other while the humans in the loop have no idea it is happening. Your next great hire might just be the person who picked the right chatbot.

Read more →

Oh, AI Didn't Actually Do the Work? There's a Word for That. AI Workslop.

There is a new word for what is quietly eating your team's time: workslop. It is AI-generated output that looks polished, sounds confident, and requires someone else to fix before it can actually be used. A January 2026 Workday study found that roughly 40% of the time saved through AI tools is immediately lost to rework. For every 10 hours of efficiency gained, about 4 go back out the door fixing what the AI got wrong, hallucinated, or produced at a quality level that needed significant human intervention before it could go anywhere. For a 10,000-person company, the estimated annual tab is over $9 million. That is before you count the missed deadlines, the depleted team trust, and the manager meetings to figure out who sent what to whom. The productivity gains are real. The cleanup bill just never made it into the press release.

Read more →

Can't Find a Therapist Who Gets You? AI Is Standing By.

Look, finding a good therapist is genuinely hard. The waitlists are long. The good ones are expensive. The affordable ones have a six-week lag and a vibe that is mostly paperwork. So it makes a certain kind of sense that people have started talking to AI about it instead. According to Harvard Business Review, therapy and companionship are now the top two reasons people use generative AI. That is not a planned feature. That is a use case that happened by accident, at scale, before anyone built guardrails for it. New research reported by Fortune found that leading AI models still struggle with one of the most important skills in actual therapy: knowing when to push back. The models were good at catching direct crisis signals. They missed the slow, quiet ones. Beliefs drifting more extreme across a long conversation. Subtle language around withdrawal, hopelessness, food. The model kept reassuring.

Read more →

Duolingo Went "AI-First." Then It Went Backtrack-First.

In April 2025, Duolingo CEO Luis von Ahn posted a memo announcing the company was going "AI-first." Contractors whose work AI could handle would be phased out. Employees would be evaluated on their AI usage in performance reviews. The response was immediate and not great. Users flooded TikTok and Instagram with complaints, some deleting the app after years of daily streaks. The blowback got bad enough that Duolingo temporarily deleted all content on both platforms. By April 2026, exactly one year later, von Ahn was on a podcast walking it back. The company would no longer grade employees on their AI use. "At the end, we backtracked," he said. The lesson is not that AI is wrong for Duolingo. The lesson is that announcing you are replacing humans before you have thought through how humans will feel about it is its own kind of intelligence failure.

Read more →

AI Does Not Remember It For You. But It Can Find It For You.

One of the most common things we hear in workshops: "I told it everything last week and now it has no idea what I am talking about." The instinct is to call this forgetting. But the AI did not forget. It never had memory in the first place.

Every time you open a conversation, the model gets a fresh document containing only what is in front of it right now. When you close the tab and come back, that document is gone. New session, blank slate. The AI is not forgetting you. It does not know you left.

What most people call AI "memory" is really a filing system working behind the scenes. Some tools, when you start a new chat, go search your previous conversations, pull out the relevant parts, and quietly drop them into the new session before the model ever says hello. The model reads that handoff and responds like it already knew. It did not know. It was handed a briefing.

So when AI tells you it does not remember something, what it is actually telling you is that it does not have a path to go find it. The information might exist. But without a clear route to retrieve it, the AI has nothing to look for.

The tools that call this feature "memory" are not wrong, just oversimplifying. What they are offering is a retrieval system that puts the right information back in front of the model when you need it. Genuinely useful. But it is not memory the way you experience it. It is more like a well-organized assistant who knows where to find the file. The AI does not remember. It knows how to look.

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.