After months in stealth, my co-founder @helloericsf and I are finally sharing @cimentoai with the world. 🌎
AI changed social engineering. Attacks are now personalized, convincing, and cheap to generate at scale.
Today we're releasing Composite-Bench: a long-horizon computer-use benchmark with certified-optimal answers and verified compute.
The strongest open-weights model isn't Kimi K3. GLM-5.2 beats it by 32 points and clears every closed model we tested except Claude.
You spent a decade training humans not to click.
Now agents click for them.
Social engineering isn't going away. It's getting a new victim.
Our founder on why the next phishing target isn't human:
we've had engineers turn down Palantir for us from schools nobody in AI is fighting over, like Utah or BYU, which helped build the early internet before GPUs were cool.
Recruit where everyone else does, and you're overpaying for an illusion, not talent.
Not needing to raise does not preclude someone from raising in the future from an incredible place of strength.
Call it the Series FU
FU money lets you walk away from anything but the absolute ideal deal.
Click rates are down. Dashboards are green. Employees still get owned.
Our CEO on why single-email phishing tests measure the wrong thing and what measuring human risk from first principles actually looks like:
https://t.co/Dj6XO8CkLs
Support tickets are where customer trust goes to die.
We run everything through shared Slack channels instead. Customers message us like teammates, engineers answer directly, and problems die in minutes... Exhibit A from today (one of the leading AI cos):