Pieter Levels's predictions
Public predictions Pieter Levels (@levelsio) made from 2023-11 to 2026-11, each checked against what happened.
Wishful thinker#4 of 5
2 partly true · 3 wrong · 7 pending
By topic
- AI 17% of 3
- Society 50% of 1
- Other 0% of 1
Scored predictions
- Partly true2025-09· AI· due 2026
After avatars (2023), photos (2024) and video (2025), 2026 would be the year of AI world models: interactive explorable 3D worlds generated from images or prompts.
World models did ship in 2026 (Google's Project Genie for AI Ultra subscribers in January, World Labs Marble 1.1 in April), but they stayed gated or niche rather than becoming the year's defining mainstream AI product the way video did in 2025.
- Partly true2025-05-03· Society
Tech hiring is dead and will not come back, as AI makes most tech jobs a thing of the past.
Indeed software-dev postings bottomed around May 2025 and then rebounded roughly 15-30%, but by late 2026 they were still about 20% below pre-pandemic levels, so hiring recovered somewhat without returning to normal.
- Wrong2024-11-01· AI
Google under Sundar Pichai had squandered its head start in AI, and this would be taught for decades as a business-school case of a fumbled lead.
Google came back hard: Gemini 3 (Nov 2025) was widely seen as taking the lead, Alphabet was the best Magnificent Seven stock of 2025 (about +65%), and Levels himself bought over $1M of Alphabet in Dec 2025.
- Wrong2024-09-27· Other
LangChain is dead
LangChain raised at a $1.25B valuation in Oct 2025 and remains widely used.
- Wrong2023-11-18· AI· due 2023-12-31
I think @sama rn is raising $10-$20B in his chat groups for his new AI company which you all predict will be called ACCELERATE Meanwhile @gdb is coding a first version of it to train their first model Investors can ...
Altman returned as OpenAI CEO within days; no new $10-20B AI company was raised.
Pending
Not resolved yet, so not counted.
- Pending2026-02-01· AI· due 2027-12-31
We're entering an exponential take off of AI reply bots on social media I now detect and block over 6 per day and almost 200 per month Last year it was just a few per week At this rate social media is going to be 9...
'By next year'. Resolve end of 2027 (hard to measure; may become UNVERIFIABLE).
- Pending2026-01-15· AI· due 2027-12-31
This took 8 minutes to generate 30 seconds of motion tracked video AI estimates it'll be real time by next year so then people will be doing live streaming like this, crazy! 😳
Relays an AI's estimate; resolve end of 2027.
- Pending2026· Politics
As Europeans leave and tax income drops, the EU will next introduce a European exit tax and then try a US-style worldwide income tax.
No EU-level exit tax or citizenship-based taxation had been enacted as of Nov 2026.
- Pending2026· Business
Big AI companies will keep absorbing what used to be standalone apps, with money flowing out of the traditional economy into AI labs.
Open-ended trend call with no deadline; too early to judge.
- Pending2025-10-11· AI
People will mostly use personal, throwaway software generated inside AI chat apps, with non-technical users not even realizing they are building software.
No deadline. Chat-app builders are growing, but most people still use conventional apps as of Nov 2026.
- Pending2025-05-03· AI· due 2030
Extrapolating a 10% yearly drop in software development jobs, almost half of today's developer jobs would be gone within five years because of AI.
Deadline is 2030. So far the trend has gone the other way: dev postings have risen since the May 2025 low.
- Pending2023-12-13· Politics· due 2028-12-13
Confident we'll get a Netflix documentary in 5 years like Painkiller, which is about Oxycontin But about the overprescription of ADHD drugs to everyone, not just in US, now in EU and rest of world too What Painkille...
5-year horizon.
Pieter Levels's maker profile →Data updated 2026-11-05
Compare with other makers
92%
44%
27%
0%
100%
100%
Frequently asked questions
How accurate are Pieter Levels's predictions?
Pieter Levels got 20% of 5 resolved predictions right (0 came true, 3 were wrong), with 7 still pending.
How does Pieter Levels rank as a predictor?
Pieter Levels ranks #4 of 5 makers with at least 3 resolved predictions on IndieMakerList.
How is the score calculated?
Came true = 1, mostly true = 0.75, partly true = 0.5, wrong = 0, averaged over resolved predictions. Pending predictions aren't scored until they resolve.