AI NEXTIA / RESEARCH OBSERVATORY
Research papers, verified developments, and thoughtful interpretation. We label company reports, independent assessments, and experimental prototypes rather than treating all announcements alike.
01 / RESEARCH SIGNALS
A curated starting collection of original sources, updated as new evidence is reviewed.
METR measured an approximately 1.9× acceleration in an Anthropic capability trend since October 2025; attributing the improvement to AI is uncertain.
Important evidence of AI-assisted R&D, not proof of autonomous recursive self-improvement.
OpenAI described a campaign yielding 948 reproduced privilege-escalation exploit examples and candidate patches requiring review.
Finding and reproducing a vulnerability is not the same as deploying a fix.
Anthropic documented AI-assisted sequence work, experimental planning, microscopy analysis, and a shortened purification workflow.
Single-team result with documented AI mistakes; independent replication remains important.
Anthropic described FlashProve, a method for verifying which AI model produced a result without exposing model weights.
Still experimental and computationally expensive.
Annual evidence collection covering capabilities, industry investment, adoption, policy, and education.
Annual snapshots require context; not a live real-time tracker.
Epoch AI tracks compute, training runs, and model development trajectories.
Data coverage varies and estimates may later be revised.
AI NEXTIA / AUTOMATED DISCOVERY
arXiv AI (cs.AI) · Unreviewed headline
Original publication ↗Research / Nov 6arXiv AI (cs.AI) · Unreviewed headline
Original publication ↗Research / Nov 6arXiv AI (cs.AI) · Unreviewed headline
Original publication ↗Research / Nov 6arXiv AI (cs.AI) · Unreviewed headline
Original publication ↗Research / Nov 6arXiv Machine Learning (cs.LG) · Unreviewed headline
Original publication ↗Research / Nov 6arXiv Machine Learning (cs.LG) · Unreviewed headline
Original publication ↗Automated external headlines, not independent editorial verification. Last checked: Nov 6, 05:17 UTC.
02 / RESEARCH FRONTIERS
A useful starting feed for trending technical papers. Trending does not equal verified.
Explore papers ↗PREPRINTSNew and recent computer science AI research, including unreviewed preprints.
Browse recent ↗HISTORICAL DATADatasets and analysis on AI progress, computing trends, and model development.
Explore datasets ↗ANNUAL SYNTHESISEvidence-driven context on industry, capabilities, and societal implications.
Read research ↗TRACK CHANGE OVER TIME
Our capability ledger records evidence type, date, availability and the question we still need answered.
