February 7, 2026
AI wrote it, devs roasted it
StrongDM's AI team build serious software without even looking at the code
Bots built a security app — $1k‑a‑day AI spend has devs fighting
TLDR: StrongDM says its AI “factory” ships security software with zero human code or reviews, tested against secret scenarios and a cloned “digital twin” of partner services. Comments split between ROI fans who love the speed and skeptics who call it token‑burning chaos and want humans in the loop.
Internet, meet chaos: StrongDM bragged that its “Software Factory” lets AI write and ship code with no humans allowed—not even for code review—and the crowd immediately split into two camps. On one side, ROI die‑hards cheered the $1,000‑per‑engineer‑per‑day token burn as the price of speed. “If it gets the same work done in five days instead of a month, the math wins,” argued one commenter, while another predicted token costs will keep dropping. Some even read the $1k line as a rite of passage: if you aren’t spending that today, you haven’t pushed hard enough yet.
On the other side: skeptics firing off “you’re just burning cash while bots fumble” warnings and insisting humans must stay in the loop. The spiciest clapback? Calling out “holdout” tests and asking if anyone actually knows what those should look like before the AI starts guessing. Meanwhile, StrongDM’s big reveal—a “Digital Twin Universe,” basically a videogame world cloning services like Okta, Jira, Slack, and Google Docs—had the thread joking it’s “The Sims for APIs,” but also nervously eyeing the fact it’s security software.
Amid the melee, a rando humble‑bragged about 40k newsletter subs and ad money via EthicalAds, because of course someone had to drop their stats. Verdict: half the internet says “future,” half says “YOLO mode with real money,” and everyone’s doomscrolling.
Key Points
- •StrongDM describes a “Software Factory” where agents generate and validate code without human writing or review.
- •The approach is justified by perceived LLM inflection points in late 2024 and November 2025 (Claude Opus 4.5, GPT‑5.2, Claude 3.5).
- •Quality is assured via scenario “holdouts” (inspired by Scenario testing) stored outside the codebase and measured by a probabilistic “satisfaction” metric.
- •A Digital Twin Universe replicates third‑party services (Okta, Jira, Slack, Google Docs/Drive/Sheets) to test APIs, edge cases, and behaviors at high scale.
- •StrongDM applies this to security-related software managing user permissions, and suggests significant token spend per engineer to improve the factory.