2026-05-17: 5 signals to keep.

A daily selection from the Strange Future Lab archive, rewritten as compact briefs with source trails.

#1 9.4 Weirdness

Companies operating under "AI psychosis" (Mitchell Hashimoto viral post).

Entire companies are now under AI psychosis—it's fine to ship bugs because the agents will fix them faster than any human can read the code.

Concrete cultural/identity breakdown among builders—entire companies losing touch with material reality as AI agents handle (or hallucinate) more of the stack; darkly funny collective delusion with real workflow implications.

What Happened

@mitchellh X/tweet (mid-May 2026) + massive concurrent HN thread (2000+ pts, heavy builder discussion on vibe-coding, shipping bugs because "agents will fix them," and roadmap-as-reality delusions). Cross-checked primary tweet and HN comments.

Source Trail

#2 8.5 Weirdness

arXiv's new 1-year ban for AI-generated "slop"/hallucinations.

Submit AI-hallucinated papers or fake references to arXiv and you're banned for a year—no more scientific slop.

Major academic institution enforcing concrete rules on what counts as legitimate human scholarly work amid flood of AI garbage; reshapes science institutions, authorship, and identity.

What Happened

Thomas Dietterich X thread + Ars Technica/404 Media coverage (May 15, 2026); explicit new moderation policy tying authorship responsibility to human verification, with one-strike penalties for un-checked LLM output (hallucinated refs, meta-prompt leftovers, illustrative tables, etc.).

Source Trail

#3 8.4 Weirdness

Patients showing plastic surgeons impossible AI-generated faces from ChatGPT (driving them "crazy").

ChatGPT is generating impossible perfect faces patients demand from surgeons—and it's breaking plastic surgery.

AI warping human beauty standards, body identity, and medical practice in absurd, concrete ways—technology changing culture and the "art" of the self.

What Happened

Business Insider reporting (recent, ~1 day old as of search) + HN/X circulation; real clinic behavior where patients prompt AI for hyper-idealized, physically impossible "after" images that reshape expectations.

Source Trail

#4 9.0 Weirdness

OpenAI + Malta government "AI for All" partnership (ChatGPT Plus for citizens).

One country just turned ChatGPT Plus into a citizen benefit—complete AI school first.

Sovereign nation-state fusing with corporate AI to make literacy a civic ritual and subsidize universal access; weird institutional future blending government, education, identity, and frontier tech.

What Happened

Official OpenAI announcement (May 16, 2026), Reuters/Engadget; citizens get 1-year free ChatGPT Plus after completing University of Malta AI literacy/responsibility course, managed by Malta Digital Innovation Authority—world's first national-scale rollout.

Source Trail

#5 7.6 Weirdness

Frontier AI breaking the open CTF format (leaderboards now measure AI orchestration).

AI didn't just win CTFs—it killed the scene. Leaderboards now reward who best deploys frontier models, not who hacks best.

Hacker/builder culture's core ritual subverted—concrete shift in what "skill," competition, and security expertise mean in an agent-dominated world.

What Happened

kabir.au blog (early-mid May) + recent HN surge, YouTube discussions, and X chatter on how models like Claude Code/GPT-5.5 + agents solve medium/hard challenges, turning scoreboards into "pay-to-win" token spend contests rather than pure human skill.

Source Trail