Narrow sample, companion material, or practitioner experience without measured outcomes.
Each source carries one line on why it got its tier. Tiers rate the evidence and the standing of the source. Whether I agree with it plays no part.
32 sources filed · EN 19 · RU 9 · ZH 4
Architecture: tools are thin adapters; a shared core handles one HTTP pool per process, API key injection, logging, tracing and error mapping.
5 of 44coffee shops returned for 'nearest coffee shop' with no point given, all in Vorkuta (latitude about 67.5)
The original hypothesis, about 18 months before the talk: cut average customer handling time by at least 30 seconds without lowering quality.
12,000active users, mostly contact-centre operators, out of 75,000+ with access
Software flaws: when a university-hosted tool returned an error, Claude Mythos Preview found a file-serving script, copied the server's code, found an injection flaw and used it to run its calculation on the server.
allcases in the report that the new detection tooling blocked when tested against them
Interface: start "task name" opens a session and returns the current SDK description; run "Python script" executes code on the server in a sandbox with the user's permissions.
2commands in the whole agent interface: start a session, run a Python script
He estimates coding-agent penetration among developers at about 99%; every other vertical is far behind.
61%pass rate of the best frontier model on Alibaba's 107 real e-commerce tasks, with no human intervention
Prime Agent, launched in August, has been downloaded more than 300,000 times and processed over 8 trillion tokens.
2,209agents in the swarm, exchanging 16,758 agent-to-agent messages and using 228.70B tokens
From running an incubation lab for Tencent in New York: in business, technology is "not the most critical or useful thing, not even the second".
3 of 10failure rate he calls unacceptable for an industrial robot policy
Lynch started with coding agents in February 2025 and expected agents to catch up with the models within six months; he says they did not.
10subtasks OpenCode created for a 1.5k-line feature, then ran one by one
A two-week iteration cycle (spec, design, build, test, release, measure) now takes a team of one or two people half a day.
30+products Guo Yu built alone with coding agents in roughly three to four months, many unreleased
By August the weekly limit ran out in about 3 days; friends hold top subscriptions to Claude and GPT at once.
~8xcost of the same actions in an 800K-token session versus a 100K-token session
IO Breakout Scale runs from 1 to 6; Russia-origin "Dark Clark" rated Category 5, Iran-origin "Bogus Bylines" Category 4 overall.
Category 5IO Breakout Scale rating of the Russia-origin 'Dark Clark' operation, the first Category 5 operation OpenAI has disrupted
Reading the VOC pages alone at two minutes a page, full-time, would take about 70 years; embedding them took one overnight run.
about $3cost for a small filter model to read 59,000 elephant mentions
Several colleagues contacted him within a day of the announcement, asking whether he would pay out the whisky; he will not.
at least threeunpublished, unrefereed, non-arXiv lecture notes cited as references in the OpenAI preprint
Earlier stages of the experiment: git plus Obsidian (8 February 2026), Personal OS structure (27 February), comparison with Karpathy's LLM Wiki (5 April), an agent persona (23 May).
~1,500documents in the Personal OS knowledge base moved from a git repo to a server
Historical comparison: railroads about 2.2% of GDP (1870-1890 peak), highways about 1.1% (1950s), telecom/fiber about 1.1%, canals about 0.66%, electrification about 0.5%.
$10.3 trillionprojected US AI data center buildout, 2025-2032
A single human playthrough takes about 2-4 hours; a high score means completing up to 21 objectives within limited turns.
100%GPT-6 Astra score on the original EBR-bench, reached in over half its attempts via one card
Prompt: a three.js (pnpm) visualization of the book's 55 cities; no questions allowed; 6 hours of work.
1 h 25 min, ~$74Claude Opus 5.5 run time and API cost, with 6 parallel subagents
Agents use two tools: search (query to documents) and fetch (URL to page text).
>80%share of GPT-5.6 Luna's search queries that use quotation marks in Artificial Analysis runs
Starting point: Lehman's second law of software evolution.
~7xreduction in cost per evolution cycle the agent achieved on its own, starting from about 30% cache hit rate
10 skills built in about two months with Anthropic's skill-creator: acceptance criteria, Bamboo deployment, WireMock stubs, task testing, OpenSearch logs, bug reports, Allure test cases, daily test report and others.
10Claude Code skills built over about two months, covering the full QA workflow
Launched 28 May 2023 with just over 200 problems; comments added August 2025.
1221problems hosted on erdosproblems.com, with over 9000 comments and almost 2000 registered users
The README lists 719 manuscripts in 372 families (the Library title says 722); a family groups a principal result with companion arguments, consequences or alternative proofs, classified by discipline.
719manuscripts in the catalogue, organised into 372 families (README at time of reading)
Hardware: Inspur YPCB-00338 card, Xilinx Kintex-7 xc7k480t, two DDR3-1066 channels with a 17.1 GB/s peak, clocked at 133.33 MHz; the host is an Intel Core i7-4790.
85.8 tok/sdecode speed for LFM2.5-230M with 4-bit weights on the FPGA card (device time)
Subscriptions are about 10% of Anthropic revenue but take over 40% of inference compute, lowering blended revenue per MW by about $36M.
~5xAPI-equivalent value of Claude plans vs OpenAI plans at the mid-tier models (Opus 5.5 vs GPT 6.1 Sol)
Median researcher daily spend: $0.17 (week ending 18 January), $10.45 (8 March), $136.68 (31 May), $601.25 (15 August).
$601.25median OpenAI researcher's daily coding-agent use at API prices, week ending 15 Aug 2026
40% of business leaders expect AI to help them get more out of existing employees; 28% expect it to reduce headcount.
-25%fall in job-posting demand for basic AI skills such as simple prompting after a January 2026 peak
Edits to Wikimedia wikis believed to come from OpenAI agents, almost all test edits in sandbox areas not visible to general readers.
millionsautomated API requests and crawled pages attributed to OpenAI-operated agents
FDE Pulse puts the median annual salary of overseas FDE roles with public pay at about $200,000 (about RMB 1.34 million); Chinese firms advertise RMB 30-50k per month.
~$200,000median annual salary of overseas FDE roles with public pay, per FDE Pulse (about RMB 1.34 million)
Codebook substitution saves about 10%; instructing models to write cablese in lowercase saves 40.4% (gemma-4-31b), 48.9% (qwen3.8-27b) and 48.4% (GLM-5.3-Flash).
40-49%fewer billed output tokens when models write records in lowercase cablese
The specification is the single source of truth; code is treated as secondary and in principle regenerable in another language.
about 90%share of reviewer blockers caused by missing edge-case tests
56.9k jobs added in September, led by Public Administration and Health Care and Social Assistance; Information and Transportation declined.
48%fall in the number of firms newly adopting generative AI from the April 2026 peak
September cuts: 43,281, down 18% from 52,881 in August and 20% from 54,064 in September 2025.
120,1362026 year-to-date job cuts attributed to AI, about 21% of all cuts and the leading reason
2 sources filed · EN 2
Structure-aware chunking is supported for nine languages (Kotlin, Java, Python, JavaScript, TypeScript, C#, PHP, Go, Rust); everything else falls back to line-based splitting.
32xstorage reduction from one-bit binary vectors versus 32-bit floats
Three scenarios: modest (impact like the internet), substantial (bigger than the internet or the railroad), extreme (likely requires recursively self-improving AI adopted quickly).
$34.1T / $36.3T / $44.4TUS GDP in 2030 at 2025 prices in the modest, substantial and extreme scenarios
2 sources filed · EN 2
Accenture's July 2026 Pulse of Change survey (n=3,000 C-suite leaders): 58% plan to change the human-AI split of work within 12 months, more than double the share at the start of 2026.
58%of 3,000 C-suite leaders plan to change how work is split between people and AI in the next 12 months
35% of organizations run agentic AI in production, 44% plan to, and more than a third expect agents to hold independent decision rights within three years (BCG and MIT Sloan Management Review study).
35%organizations already running agentic AI in production (BCG and MIT Sloan Management Review)
2 sources filed · EN 2
Most large companies are not actively scaling agents, and those that are report limited bottom-line effect (citing unnamed recent surveys).
30%+uplift in wealth-advisor revenue productivity at a Southeast Asian bank (BCG client claim)
At the end of 2025, AI models completed tasks taking humans three to four hours with a 65% success rate (Mertens et al.).
65%AI success rate at end-2025 on tasks taking humans three to four hours
2 sources filed · EN 2
Four futures: "colleague machine" (labor, shared gains), "output engine" (infrastructure, shared gains), "hollowed middle" (infrastructure, few winners), "first-mover future" (labor, few winners).
93% / 7%share of organizational AI investment going to technology infrastructure versus work and people issues (Deloitte study)
The fluid starts at rest, a smooth force is applied, energy stays finite, and speed grows without bound: statements "C" and "D" of the official problem.
~10,000concurrent agents in the group that produced the Navier-Stokes resolution
3 sources filed · EN 3
No widespread, economy-wide job displacement associated with AI.
19%employment shortfall for workers aged 22-25 in highly AI-exposed occupations versus less-exposed peers, June 2026
Deployment stage: 42% are testing or have a few deployments, 43% are expanding across functions, 15% have scaled orchestrated multi-agent adoption.
31% → 74%leaders expecting at least half of processes redesigned around agents within two years vs four years
A task vulnerability index built from eight task traits: digital observability, output regularity, ex-post evaluability, retrieval intensity and error tolerability push exposure up; tacit context, liability and coordination-and-trust push it down.
19 ppdrop in correct solutions on a task outside the AI frontier in the 758-consultant field experiment the paper cites
1 source filed · EN 1
AI use among firms in OECD countries rose from around 7% to 20% between 2021 and 2025, with larger firms and start-ups ahead and SMEs held back by cost, infrastructure and skills shortages.
7% to 20%share of firms in OECD countries using AI, 2021 to 2025
2 sources filed · EN 2
Leaders feel less prepared to manage disruption than earlier in 2026, down seven percentage points on average.
23%C-suite leaders reporting widespread, sustained business value from AI, down from 32% earlier in 2026
Only 17% of organisations report advanced or transformational workforce readiness, defined as AI embedded in workflows and linked to measurable business outcomes.
3%organisations reporting leaders highly prepared to manage AI-enabled ways of working
2 sources filed · EN 2
Worker access to AI tools grew 50% in one year, yet fewer than 60% of those with access use AI daily and 84% of organisations have not redesigned jobs or workflows around it.
84%organisations that have not redesigned jobs or workflows around AI (Deloitte State of AI in the Enterprise 2026)
11% of leaders place their organisation in reinvention; nearly 90% remain in enablement or automation.
11%leaders who say their organisation is in the reinvention horizon
2 sources filed · EN 2
More than 90% of organizations use AI in talent acquisition, and fewer than 5% report transformational outcomes across most metrics.
54%of organizations say AI is making candidate evaluation harder
It draws on RAND's 2024 model-weights work, which maps five security levels (SL1-SL5) to five threat-actor capacities (OC1 amateurs to OC5 top nation-states).
5security clusters: access and interfaces, development and supply chain, monitoring and response, personnel, physical
1 source filed · EN 1
AI specialist postings rose 68.9% from 2024 to 2025, while all postings rose 8.6%.
68.9% vs 8.6%growth in AI specialist job postings vs all job postings, 2024 to 2025
3 sources filed · EN 3
A physical-activity coaching case (GPTCoach and the Bloom app): expert coding found over 93% of the coach's conversational strategies were consistent with motivational interviewing or neutral, though gaps against skilled human coaches remained.
5xincrease in app usage time with the LLM coach versus a no-LLM control, while physical activity stayed comparable (N=54, four weeks)
Students using AI for homework rose from 48% (May 2025) to 62% (December 2025), while 67% say it may harm their critical thinking (RAND).
48% to 62%students using AI for homework help, May to December 2025 (RAND, cited)
Countries covered: Australia, Canada, France, Germany, India, Italy, Japan, Singapore, the UK and the US.
71crossover points where different teams in one agency asked for the same or similar assurance information
5 sources filed · EN 5
Principle 1, strategic oversight: boards should balance short-term productivity pressure against sizeable early investment in data, infrastructure and change management whose returns are uncertain, and consider effects on business model, partnerships and M&A.
5governance principles, broken into 22 sub-principles
86% of organizations plan to increase AI investment in 2026, but only 21% report redesigning end-to-end processes with AI at the core (Pulse of Change survey: 3,650 executives, 20 industries, 20 countries).
21%executives reporting end-to-end process redesign with AI at the core (Pulse of Change survey, 3,650 executives)
Turn 1 scenario: a fictional US firm releases an AGI-level model; GDP rises while white-collar unemployment and protests grow.
Executives surprised by political risk events "most or all of the time" rose from 1% in 2021 to 35% in 2025 (EY-Parthenon Geostrategy in Practice).
1% to 35%executives surprised by political risk events 'most or all of the time', 2021 vs 2025 (EY-Parthenon)
Participants by location: UK 23, EU 11, USA 9, Canada 6, Hong Kong 4, Australia 4, Japan 3, other Asia 2.
62 / 60institutional investors in six roundtables / interviews, from 12 countries
5 sources filed · EN 5
Computing cost fell about seven trillion-fold between 1940 and 2006 (Nordhaus); AI differs from traditional computing by handling tacit knowledge, in Polanyi's sense of "we know more than we can tell".
0.548 SD to 0.139 SDeducation-based productivity gap without vs with AI access (Cruces et al. 2026, N=1,174)
Bain's AI readiness survey: 80% of generative AI use cases met or exceeded expectations, yet only 23% of companies can tie initiatives to measurable revenue gains or cost reductions.
80%generative AI use cases that met or exceeded expectations in Bain's AI readiness survey
Only about 15% of organisations use AI to fundamentally redesign how work is performed (citing Accenture, 2025).
~15%organisations using AI to fundamentally redesign how work is performed
The US "Winning the Race: America's AI Action Plan" (2025) treats open models as having "geostrategic value"; China's "Global AI Governance Action Plan" frames open source as global sharing aimed at the Global South.
nearly 1,000%growth in DeepSeek model downloads on model-sharing platforms since R1's release in January 2025 (US CAISI)
Five overarching recommendations: scale governance with degrees of autonomy, keep human control and accountability, monitor continuously after deployment, use defense-in-depth and containment, and assess risk at system level rather than model level.