Useful AI agents by September or October 2025
Details
Partial update to my 'useful agents' timeline based on Operator release
Evidence
- Jan 23, 2025: Operator released but underwhelming for demoed use cases
I will not be regularly using codex-cli in 3-6 months
Details
I just think there will be something better
SourceWe'll have some version of actually impactful AI agents before the end of 2025
Details
Test time compute paradigm of o1 series makes this even more likely. 'Meaningfully impact our lives' is vague, but we'll know it when we see it.
Evidence
The first industry that will be hugely disrupted is customer support agents in call centers
Details
At least one firm will do major layoffs in this area and replace/supplement workers with AI systems. Customer satisfaction will go up not down.
SourceAt least one, but certainly not the majority of major tech companies dramatically changes hiring methodology in light of AI
Details
For example, allowing the use of models during an interview, but having higher expectations
Evidence
- Jun 15, 2025: No major announcements yet, checking in at halfway point
A jarring step-function level release sometime later this year
Details
Perhaps GPT5, perhaps an agent, maybe whatever is next from Anthropic
SourceAI will reach top 50 worldwide at competitive coding with next release
Details
o3/o4-mini already top 200. Similar 800 point jump would put next release at ~3500 ELO
SourceBy 2026 AI will be pervasive enough that the predominant way of using AI will not be through chat/co-pilots
Details
People will still think primarily of chat interfaces as AI, but this will begin shifting in 2025
SourceThere will be a widely popular AI generated song
Details
There's just too much latent creativity in the world that is enabled by these tools
SourceUBI trials in the US will continue to show disappointing results
Details
Less optimistic than I once was about this as a path forward
SourceAI task horizon will reach ~32 hours (work week) with just 3 doublings in 2 years
Details
Based on METR data showing doubling of task length capability
Source10 doublings in AI task capability over next 2 years (90th percentile case)
Details
Doublings every 3 months for next year, then every 2 months following year. Would mean AI handles year-long tasks.
SourcePeak AI anxiety will occur in 2026 or 2027 (not 2025)
Details
Normal people will become increasingly concerned about whether it will impact their industry and jobs
SourceMajority of current intermediate-level knowledge work will be automatable
Details
15th-85th percentile understanding will be increasingly done by AI
SourceAI video generation as creative hobby possible in 3-5 years
Details
Updated timeline after Sora disappointment
Evidence
- Dec 31, 2024: Sora is just not that good yet. Updated timelines from 2025 to 3-5 years