Interesting that TBL identifies exactly 6 platforms as "harmful" (Facebook, Instagram, Snapchat, TikTok, X, YouTube), but the solution of personal data wallets doesn't address why those platforms are harmful. Centralizing user data in pods doesn't solve addictive algorithms or misinformation; those are business model problems, not technical ones.
This isn't AI replacing workers - it's using AI as cover for offshoring and wage suppression.
The playbook: Announce layoffs citing "AI efficiency gains" (stock jumps), quietly rehire 6 months later at half the cost offshore, then claim "AI productivity" let you do more with fewer FTEs.
The tell: If AI actually replaced the work, you wouldn't need to rehire at all. The fact they're rehiring at lower wages proves the work still needs humans.
The best architecture decision is the one that's still maintainable when the person who championed it leaves. Always pretend the person who maintains a project after you knows where you live and all that.
The evaluation framework might be the real moat here. Coding agents work because "tests pass" is an instant signal, but what about domains where success is subjective or delayed by months?
Medical diagnosis? Legal research? Customer support quality? I wonder if agent labs in these types of domains are struggling to improve as fast, even with great workflow data.
