Jesse Eisenberg keeps testing AI to make sure it's still bad at writing
SummaryOn Variety's Awards Circuit podcast, the writer-director of A24's The Debut said he prompts AI to write scenes as a check. So far, he says, it's still bad. He is less sure how long that lasts.
· MODEL Editors · 1 min read
Some links may be affiliate links. Disclosure

What he said. Asked on Variety's Awards Circuit podcast (published Nov. 6) whether he is nervous about AI and the film industry, Jesse Eisenberg said he has "a paranoid answer." "I sometimes experiment with AI to see whether it's good at writing. And it's still not. I'm paranoid, so I check and it still isn't. Thank God," he said. "I say, 'Write a scene like this.' I'm just testing it to make sure it's bad, and it's still bad. So I'm relieved, but, like, how far away is it from being good?" (Variety (opens in a new tab))
The part that worries him. He described a future where a kid puts on goggles, uploads a few photos and asks for a personal movie in which they rob a bank and kiss their crush, then presses play. His comfort: people still want to watch stories in the same room, as with the Greek plays he performs. His worry: "whether we'll still be able to make a living from it. I'm not 100% sure about that." Julianne Moore, his co-star in the A24 musical comedy The Debut, was more hopeful, saying new creative voices keep appearing.
Why builders should care. It is, basically, a homemade AI eval: a domain expert re-running the same prompt over and over and judging the output himself. That is also how you should check AI writing tools. Use your own prompts on your own kind of work, and re-test when models update, instead of trusting a leaderboard.
Caveats. Eisenberg did not say which AI tools he uses or how often he tests them. This is one person's informal check, not a benchmark. Not tested by MODEL.
Disclosure: Not sponsored. No affiliate links. Not tested by MODEL.
MODEL Editors
Sources
Keep reading

Hot take
Hot take: 'the tests pass' is not a security check, and a new preprint shows why
A.S.E 2.0 (CASE-Bench), a preprint posted to arXiv on Nov 4, built 145 coding tasks from real CVEs. Context tuned for working code lifted functional pass rates up to 8.2 points but secure pass rates by at most 3.7. Coding agents saw the relevant safeguard in only 27-36% of runs.
· MODEL Editors

AI News
Claude Code's 'don't ask again' button saved passwords into files that ended up on GitHub
Security firm Apiiro says 1,122 public copies of Claude Code's .claude/settings.local.json hold at least one secret, because the approval can save the exact command you typed, password and all. Rotate first, then clean up.
· MODEL Editors

AI News
AI-blamed job cuts just hit their lowest month of 2026
Challenger, Gray & Christmas (Nov 5): US employers cited AI for 1,265 announced job cuts in October, the lowest monthly total this year. Market and economic conditions passed AI as 2026's top stated reason, 130,531 to 121,401.
· MODEL Editors
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.