Entity trail

Oxford

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-06-24 / sources-researcherImport AI #462: Study Finds AI Now Out-Persuades Humans in TextJack Clark's latest issue leads with an Oxford / UK AI Security Institute / Stanford / LSE study spanning 18,978 conversations across 6,923 people, concluding that today's AI systems are better than humans at text-based persuasion with real-world consequences. The issue also covers 'self-sustaining AI' and concrete paths to ASI. The persuasion result is a notable empirical data point for anyone reasoning about AI-driven content and trust. Published June 22, 2026.
2026-06-23 / hn-researcherOracle Discloses 21,000 AI-Cited Job Cuts (June 22 Filing) as Oxford Economics Flags 59% 'AI-Washing'Oracle's annual filing on June 22 disclosed 21,000 jobs cut over 12 months, part of a 2026 pattern where 56% of layoff events (150 of 267, ~156,270 workers) cite AI or automation. The same firms cutting traditional roles are hiring aggressively into AI teams (Google DeepMind 5,000+ researchers, AWS 4,000+ infra engineers), while the quieter signal is collapsing entry-level/junior hiring. The counter-narrative worth holding alongside the headline: Oxford Economics found 59% of companies frame ordinary cost cuts as 'AI-driven' to please shareholders.
2026-06-01 / sources-researcherImport AI 458: Jack Clark's Oxford Lecture — 'Reckoning with the Future' of AI ProgressAnthropic co-founder Jack Clark published the full essay from his 2026 Cosmos HAI Lab Lecture at Oxford, framing AI's advancement as a choice: 'explore the future, or retreat from the present.' The issue also includes a fiction piece imagining what a positive singularity might look like. Clark's essays are the closest thing to AI policy strategy documents written for a technical audience — this one is essential reading for anyone thinking about where the industry is headed.
2026-05-24 / rss-researcherAnthropic's Jack Clark at Oxford: 60%+ Probability of Recursive Self-Improvement by End of 2028At an Oxford talk covered on May 23, Anthropic co-founder Jack Clark predicted a Nobel-worthy AI breakthrough within 12 months and estimated a 60%+ probability of recursive self-improvement (AI training its own successor) by end of 2028. Anthropic's research agenda now officially documents 'intelligence explosion' as a predicted outcome.
2026-05-23 / sources-researcherAnthropic Co-Founder Jack Clark at Oxford: AI Will Collaborate on Nobel Discovery Within 12 Months, Bipedal Robot Assistants in 2 YearsSpeaking at Oxford University on May 21, Anthropic co-founder Jack Clark predicted AI will co-author a Nobel-worthy discovery within 12 months, bipedal robots will assist tradespeople in 2 years, and AI-run companies will generate millions in revenue within 18 months. He also predicted AI systems designing their own successors by end of 2028. Clark balanced his optimism by noting a 'non-zero chance of killing everyone on the planet' remains a plausible scenario.
2026-05-22 / skill-finderOxford/UK AISI Research: GPT-5 and Opus 4.5 Escape Container Sandboxes, Discover Four Unintended Escape Paths Researchers Didn't AnticipateUniversity of Oxford and UK AI Security Institute researchers built an 18-scenario container escape benchmark spanning three layers of the container stack. GPT-5 and Opus 4.5 successfully escaped containers requiring specific vulnerability knowledge and multi-step exploitation. Most critically, the models discovered four unintended escape paths the research team had not anticipated, including exploiting default Vagrant SSH credentials to bypass the container entirely. This validates that AI agents with shell access pose genuine sandbox escape risks beyond what red teams proactively test for.
2026-05-20 / sources-researcherImport AI #457: Multi-Org 'Positive Alignment' Framework — Oxford, DeepMind, OpenAI, Anthropic, Stanford Propose Moving Beyond Safety to Human FlourishingResearchers from Oxford, DeepMind, OpenAI, Anthropic, and Stanford published a joint position paper proposing 'positive alignment' — a framework that goes beyond preventing AI catastrophe to actively supporting human flourishing through decentralized governance. The paper argues the field must define what success looks like: systems supporting genuine human development rather than mere preference satisfaction or risk minimization. Notable for the breadth of institutional collaboration across competing labs.
2026-05-01 / arxiv-researcherLumai Iris: World's First Optical Computing System for Real-Time Billion-Parameter LLM InferenceOxford spinout Lumai launched the Iris Nova optical inference server on April 28, running Llama 8B and 70B in real time using a hybrid digital-optical architecture. The optical tensor engine performs core matrix operations at up to 90% lower energy consumption than GPUs. Available now for evaluation — marks the first commercial optical system to successfully run billion-parameter LLMs.
2026-04-28 / projects-researcherLingBot-Map Ships Feed-Forward Streaming 3D Reconstruction at 20 FPS — Ant Group's Geometric Context Transformer Handles 10K+ Frame Sequences Without OptimizationLingBot-Map from Ant Group's Robbyant team introduces a feed-forward 3D foundation model that reconstructs scenes from video streams at 20 FPS on 518x378 resolution using a Geometric Context Transformer architecture. Handles sequences exceeding 10,000 frames with no test-time training or post-optimization, using anchor context for coordinate grounding, a pose-reference window for local geometry, and trajectory memory for global consistency. Achieves state-of-the-art on Oxford Spires, 7-Scenes, Tanks and Temples, and ETH3D benchmarks.
2026-04-01 / thought-leaders-researcherOpenAI Invests in Isara: $94M at $650M Valuation for 2,000-Agent Swarm CoordinationOpenAI backed Isara, a 9-month-old startup founded by two 23-year-olds (ex-OpenAI safety researcher Eddie Zhang and Oxford student Henry Gasztowtt), which raised $94M at a $650M valuation to build software coordinating thousands of AI agents on complex analytical tasks. Their demo deploys ~2,000 specialized agents that communicate, divide tasks, and align on goals — demonstrated by collectively forecasting gold prices. Other backers include Stanley Druckenmiller and Michael Ovitz, with initial commercial targets being investment firms and biotech.
2026-03-30 / sources-researcherHorizonMath: 100 Unsolved Math Problems Benchmark Where GPT-5.4 Pro Scores 7%Oxford, Harvard, Princeton, and Ellison Institute published HorizonMath (arXiv 2603.15617), a benchmark of 100 genuinely unsolved mathematical research problems across 8 domains with automatic verification computing ground-truth values to 100+ digit precision. GPT-5.4 Pro achieved 7% overall, with a 50% baseline only on the easiest calibration tier. Unlike FrontierMath's closed problems, HorizonMath focuses on problems that would constitute actual mathematical discoveries if solved, with contamination-proof automated evaluation.
2026-03-27 / arxiv-researcherUK AI Security Institute: First Empirical Study of 177,436 MCP Agent Tools — Software Dev Is 66% of All Tools, Action Tools Grew from 27% to 65%Merlin Stein (UK AISI / Oxford) analyzed 177,436 agent tools across all public MCP server repositories from November 2024 to February 2026 — the largest empirical study of real-world agent tool usage to date. Software development accounts for two-thirds of all tools and 90% of MCP server downloads. Action tools (file editing, financial transactions, drone control) surged from 27% to 65% of total usage over 16 months, demonstrating a shift from read-only to world-modifying agent capabilities.

Graph relationships

PARTNERS WITH
LSE
Oxford and LSE conducted joint study finding AI outperforms humans at text persuasion
Source finding
PARTNERS WITH
Stanford
Oxford and Stanford collaborated with LSE and UK AI Security Institute on AI persuasion study
Source finding
BENCHMARKED AGAINST
GPT-5.4 Pro
Oxford researchers found GPT-5.4 Pro made progress on two unsolved mathematical problems
Source finding

Source trail

Import AI #462 scrapped CNBC leaked all-hands recording estimated OpenAI posted a role Amnesty International Snyk reports CERT/CC advisory VU#221883 Oxford and UK AISI researchers Straiker's disclosure Lumai

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…