I took this very interesting list from Greg Isenberg ( https://x.com/gregisenberg/status/1956703650744578091?t=92GfqA1cXG6azEFXgxU-1A&s=19 )
and asked AI to analyze which predictions are less likely and which ones are practically already here.
Ranking of AI predictions that are already among us:
#1 AI girlfriends/boyfriends → Apps like Replika, Character.AI and similar are already making tens of millions. The “black market” in app store rankings confirms the trend.
#2 Knowing when NOT to use AI → In consulting, art, education, and journalism, the “human touch” already sells at a higher price.
#4 AI better than accountants/doctors in certain tasks → In taxes and image-based diagnosis, AI already outperforms the human average.
#6 AI therapy → Replika, Woebot and others show surprising levels of engagement, especially thanks to 24/7 availability.
#8 Local-first AI → On-device models like Gemini Nano, LLaMA optimized for mobile, or Copilot in Windows are already a reality.
#9 Founders without coding skills → No-code platforms like Bubble, GPTs, and agents already make this possible.
#21 AI killing the résumé → Companies are starting to evaluate digital portfolios and online footprints, using AI for screening.
#33 Universal translation → Near real-time is already in Zoom, Teams, and Pixel. Not perfect, but usable.
#34 Small businesses with AI → Shopify, Wix and others already integrate AI to compete better.
#39 “We’ve always done it this way” → Industries like insurance, banking, and healthcare already see AI handling bureaucracy better than humans.
On the frontier (early, but still a long way to go):
#3 Agents hiring agents → Prototypes exist (AutoGPT, Devin), but reliability is lacking.
#5 Reputation systems for agents → A “Moody’s for AI” doesn’t exist yet, but it’s logical it will emerge.
#7 $100M company with zero human employees → Some indie hackers are getting closer, but no case at that scale yet.
#11 Emails negotiating with each other → Intelligent inbox pilots exist, but not widespread.
#13 Agents creating companies → Legal and regulatory barriers are the main obstacle. Could happen in crypto-friendly jurisdictions.
#18 Chief AI Officer being an AI → Some companies are already placing AI on boards as experiments, but not at scale.
#22 First AI agent canceled on social media → Probable and close, but no iconic case yet.
#27 AI detecting AI at 99.9% accuracy → Detectors exist, but nowhere near that precision.
#30 SaaS flipping with AI → Happens occasionally, but not mainstream yet.
#35 Retirees becoming millionaires thanks to AI → Courses and packaged know-how exist, but not at large scale.
#37 Double careers → Emerging, but autonomous management is still very limited.
More fantastical or harder to achieve:
#10 AI agent losing $20M due to manipulation by another agent (before 2026) → Possible, but seeing such a large, documented case so soon feels exaggerated.
#12 End of the concept of “original work” → Authorship still matters legally and culturally. Total collapse will take much longer.
#14 AI cloning Joe Rogan and him losing the lawsuit → Legally complex; unlikely he would lose such a case this soon.
#15 Anti-AI movement bigger than pro-AI → Resistance exists, but surpassing global pro-AI sentiment is debatable.
#16 Kids being AI-first, unemployable but dominating entrepreneurship → Possible, but requires an entire generation.
#17 “Human-only” schools at $50k/year → Maybe, but that’s an elitist niche, hard to scale.
#24 World tour of a dead celebrity with new AI music → Technically feasible (ABBA already did holograms), but grossing $500M with “new songs” is still more science fiction.
#25 First universal AI-human language → A totally new language adopted worldwide is among the least likely.
#26 AI agent committing insider trading spontaneously → Very sci-fi, though not impossible long term.
#28 Biggest-ever data leak caused by a rogue agent → Feasible, but surpassing all historical leaks so soon sounds like a movie script.
#29 End of passwords, captchas, and 2FA → They are on the way out, but their total disappearance will take much longer.