“Software brain” – mapping everything to controllable databases – explains why tech insiders diagnose AI’s mass unpopularity as a $200M marketing problem when it is a product-experience problem.
Key Takeaways
NBC poll: AI has worse favorability than ICE; Quinnipiac finds 50%+ expect net harm; only 35% of Americans feel excited about AI.
Gen Z uses AI most and dislikes it most: Gallup shows 18% hopeful (down from 27%), 31% angry (up from 22%) year-over-year.
ChatGPT at 900M weekly users disproves the exposure gap theory; Sam Altman’s $200M TBPN podcast spend treats a product problem as a messaging problem.
DOGE’s database-first government takeover failed because databases diverge from reality; the author argues AI builders keep tweaking the database to match the world, not the reverse.
Ex-Michigan Supreme Court Chief Justice Bridget McCormack’s automated AI arbitration pitch assumes people accept worse outcomes from machines if they feel heard – pure software brain applied to law.
Hacker News Comment Review
HN consensus largely rejects the article’s framing: commenters want aggressive automation of drudgery, scientific research bottlenecks, and physical infrastructure – the objection is unreliability, not automation itself.
The actionable builder signal from comments: automation that requires as much verification work as doing the task manually provides zero leverage; reliability is the actual product bar, not novelty.
Several commenters call the argument a strawman – conflating recent AI-specific backlash with automation broadly, and ignoring that dishwashers, washing machines, and 90M ChatGPT users show demand is real when the tool works.
Notable Comments
@fmajid: “they want the drudgery to be done reliably and not require as much work checking as it would have doing it in the first place” – frames the reliability bar precisely.
@pizzly: argues for automating the full scientific research loop – drug candidates, lab experiments, analysis – where the limit is scientist count, not appetite for automation.
