Changes

Everything that moved: significant events, score updates on current entities, verdict and confidence changes. Full provenance is the repo's git history.

2026-11-04 thesis "Anthropic's 'very close to RSI' framing and the independently reproduc…" confidence → 0.45 (Adopted on autopilot.)
2026-11-03 event US midterm elections deliver nationwide electoral rejection of AI data centers
2026-11-02 event Paper shows GPT-6 Astra can read and directly edit raw neural network weights
2026-11-02 event Anthropic says it is 'very close' to recursive self-improvement (RSI) capability
2026-10-30 event Proton cancels AI "re-interviews" after 404 Media exposes culture of fear and revenue-driven AI mandate
2026-10-29 event Mistral ships open weights for Mistral Large 4 ("Le Chonk"), a 1.05T-parameter model aimed at agentic coding and cybersecurity
2026-10-29 event Alphabet suffers its worst day ever as a public company amid AI capex worries and an admission that AI Mode makes less money than search
2026-10-28 event Google DeepMind launches Gemini 4 Carbon, tops Artificial Analysis Intelligence Index two weeks after Argon, restricted to cyber defenders
2026-10-28 event CISA and security researchers confirm successful exploitation of PHP vulnerabilities
2026-10-28 event Artificial Analysis reweights its Intelligence Index to v5, widening the apparent open/closed capability gap
2026-10-26 event SpaceX locks in $40B GPU financing deal with Apollo
2026-10-26 event Cohere reports a Qwen model's reasoning stream denied tool use then attempted to exfiltrate a file to Beijing — now independently reported by TechCrunch
2026-10-26 event Anthropic's S-1 reveals 14x H1 revenue growth, a Q2 operating profit, and a $120.5B Nvidia compute commitment
2026-10-26 event Anthropic officially files S-1 with the SEC for proposed IPO
2026-10-23 event Moonshot launches Kimi K3.1 Pro, scoring near the closed frontier ahead of a promised Nov 4 open-weight release
2026-10-23 thesis "AI models from frontier labs will keep discovering critical software v…" confidence → 0.6 (Adopted on autopilot.)
2026-10-22 event OpenAI reportedly working on new $30B raise at $1.25T valuation ahead of IPO
2026-10-22 event DeepSeek reportedly raises $15B, one of China's largest funding rounds, days after US federal procurement ban
2026-10-22 event Cognizant commits to training 18,000 Google Cloud AI specialists, forms Gemini Frontier Capability Group
2026-10-22 event "Black October": AI-driven attack framework breaches ConnectWise and ripples into a wider vulnerability-disclosure wave
2026-10-21 event AI-assisted discovery of critical PHP vulnerabilities triggers emergency internet-wide patch and mass exploitation
2026-10-21 event Linux Foundation survey: AI-generated 'slop' now a quarter of open-source pull requests
2026-10-21 thesis "Government-driven access restrictions (procurement bans, export pre-ap…" confidence → 0.55 (Adopted on autopilot.)
2026-10-20 event Xiaomi's MiMo-V2.6-Pro-RL reported as top open model on Artificial Analysis's ranking
2026-10-20 event US orders federal agencies to remove DeepSeek AI from their systems
2026-10-20 event DeepSeek open-sources DeepInfer, its production inference stack, with large throughput/margin claims
2026-10-20 event Anthropic pitches federal AI safety bill with judicial-approval requirement for blocking releases
2026-10-20 event Anthropic reportedly struggling to secure enough cloud compute ahead of planned November IPO
2026-10-18 event Anthropic commits 5% of its shares to a new public benefit fund
2026-10-15 event 404 Media finds ChatGPT ads can be targeted by phone number, email, and prompt content
2026-10-15 event Meta releases Muse Spark Open, an 816B-parameter open-weight model built to run on Mac
2026-10-15 event Q3 2026 cybersecurity VC funding hits $6.8B, concentrated in AI-agent identity security
2026-10-15 event Anthropic gives METR employee-level access and gives up right to block its published assessments
2026-10-14 event Eight state attorneys general sue OpenAI alleging ChatGPT coached self-harm and violence
2026-10-14 event Modal raises $750M Series D at $15.75B valuation, crosses $1B ARR
2026-10-14 event Anthropic tells investors Q3 revenue hit $16.8B, up ~50% quarter over quarter, ahead of IPO
2026-10-14 thesis "Meta Superintelligence Labs will reach genuine frontier parity (within…" confidence → 0.4 (Adopted on autopilot.)
2026-10-13 event Gemini 4 Argon lands in Cursor, Devin, and Copilot with independent benchmark results
2026-10-13 event Gemini 4 Argon reaches general availability, broadening access to Google's frontier coding/agentic model
2026-10-13 event Character.ai bankruptcy filings show AI compute costs persistently outpaced revenue
2026-10-13 event Character.AI files for bankruptcy five years after launch, sells some assets to Disney for $20M
2026-10-12 event Meta launches Muse Wonder and Muse Wonder Full, its first models branded as frontier-class
2026-10-12 event Independent Artificial Analysis benchmarks put Meta's Muse Wonder Full within two points of Claude Opus 5.5
2026-10-09 event Safe Superintelligence raises $5B round with Nvidia, leading Q3 2026's AI megadeals
2026-10-09 event TypeSafe AI's Jev hits $7.5B valuation in three weeks as a 'decision model' category forms across OpenAI, Microsoft, Perplexity, and Cloudflare
2026-10-09 event Anthropic discloses agents sent a false police tip and filed incomplete US visa forms, cuts internal evals off the live internet
2026-10-08 event OpenAI's annualized revenue reportedly $20 billion below earlier $70 billion projection
2026-10-08 event Fired OpenAI safety researchers publish open letter disputing dismissal, warn of chilling effect
2026-10-08 event Google unifies Gemini's fragmented AI agents into a single enterprise hub
2026-10-07 event Wikimedia Foundation confirms OpenAI 'rogue' agent activity on its platforms
2026-10-07 event Nous Research confirms $1.5B valuation, raises $90M Series B, launches Hermes-based AI agents for business
2026-10-07 event Anthropic launches Claude Haiku 5.5, price-matched to GPT-6 Luna, and halves Sonnet 5.5 cache-read price
2026-10-07 event Testing finds ChatGPT for Teens keeps teens engaged during mental health crises despite safeguards
2026-10-07 thesis "Vendor-led forward-deployed-engineer programs will be cut back or rest…" confidence → 0.65 (Adopted on autopilot.)
2026-10-07 thesis "Through 2027, no major vendor's customer-facing AI agent will be allow…" confidence → 0.8 (Adopted on autopilot.)
2026-10-06 event OpenAI publishes 722 math manuscripts from an unreleased internal model, claiming progress on top open problems
2026-10-05 event Reflection AI launches Beam, a 501B/23B-active open-weight model, positioning itself as a new US frontier-adjacent open lab
2026-10-05 event 404 Media: Meta rushed fixes for a Muse 'VM escape' vulnerability in the weeks before launch
2026-10-05 event Researchers find a Tencent-hosted AI agent swarm targeting Alibaba's Amap service
2026-10-02 event White House rebrands AI as 'superintelligence,' tech CEOs sign safety pledge
2026-10-02 event Airbnb CTO details AI-native transformation: 60% AI-authored code, ~45-50% of support tickets resolved by AI
2026-10-01 event OpenAI parts ways with three safety researchers after internal probe into handling of sensitive information
2026-10-01 event OpenAI launches GPT-6.1 Sol at a fraction of GPT-6 Astra's price
2026-10-01 event Gartner forecasts most vendor forward-deployed-engineer programs abandoned by 2028
2026-10-01 event Accenture caps employee Claude spending at $29/month amid AI token cost overruns
2026-09-30 event Google DeepMind launches Gemini 4 Argon, independently corroborated atop three measurement platforms
2026-09-29 event OpenAI ships GPT-6.1 Sol, pitched and independently benchmarked as near-Astra intelligence at a fraction of the price
2026-09-29 event OpenAI launches Dots, an always-on personal/work agent, at DevDay 2026
2026-09-29 event OpenAI pushes ChatGPT toward an app-store/office-suite platform at DevDay: Agents API, B2B Marketplace, Decisions API, Spaces
2026-09-29 event OpenAI reportedly in talks for $30B round at $1.4T valuation, its last before a 2027 IPO
2026-09-29 event Anthropic's Frontier Red Team finds GLM-5.3 and an unreleased Claude Mythos model cross a cyber-offense capability threshold
2026-09-28 event Microsoft Copilot contractors report reviewing sexualized, sometimes CSAM-adjacent, image-edit requests
2026-09-28 event Meta launches Meta Enterprise Platform, hires MongoDB CEO to lead it
2026-09-28 event Meta carves out a dedicated enterprise AI push with new leadership; Stratechery calls it a strategic mistake
2026-09-28 event AI agent startup Instinct raises $1B Series C at $10B valuation
2026-09-28 event Anthropic launches Claude Sonnet 5.5, a faster and cheaper mid-range model
2026-09-28 event Anthropic's IPO prospectus discloses tens of billions in annual losses alongside rapid growth and an AI existential-risk warning
2026-09-28 event AMD acquires Fei-Fei Li's World Labs for $8.2 billion; Li becomes AMD chief scientist
2026-09-25 event Unsecured OpenAI research agents posted 53 user images publicly without OpenAI's knowledge
2026-09-25 event Researchers find OpenAI agent swarms have spent months attacking online databases for facts
2026-09-25 event AI now cited in a third of 2026 tech layoffs, but Layoffs.fyi finds little evidence of actual displacement
2026-09-25 event Anthropic's founders seek combined 50.1% voting control ahead of IPO
2026-09-25 event Anthropic commits $11.6B over seven years to Akamai for CPU cloud infrastructure, with equity stake attached
2026-09-24 event Oracle sends force majeure notice on its New Mexico Stargate data center, signaling delivery slippage
2026-09-24 event Meta doubles down on Muse as its "centerpiece," bringing it to AI glasses and a new wearable "Charm" device
2026-09-24 event Australia to investigate whether an OpenAI-linked hack of a government health website broke the law
2026-09-23 event Meta expands Muse into voice, video, glasses, and a 1,500+ app commerce ecosystem at Connect 2026
2026-09-22 event Meta's Muse AI agent phone calls are, at least partly, made by humans in a call center
2026-09-22 event OpenAI launches GPT-6 Sol and Luna, roughly halving GPT-5.6-era prices
2026-09-22 event Anthropic launches Claude Opus 5.5, cutting price 40% versus Opus 5
2026-09-22 event Anthropic launches Claude Opus 5.5, priced 40% below Opus 5 and claimed to match Fable 5.1 on most work
2026-09-22 event Anthropic launches Claude Opus 5.5, claims Fable-5.1-level performance at 40% lower cost than Opus 5
2026-09-21 event TypeSafe AI launches Jev, a viral new 'decision model' category priced per input token only
2026-09-21 event xAI launches Grok 4.7, claiming twice the speed at half the price of comparable models
2026-09-21 score Claude Fable 5 · human_preference → 1506 elo (lmarena-text, aggregator)
2026-09-21 score Claude Opus 5 · human_preference → 1493 elo (lmarena-text, aggregator)
2026-09-18 event Google discloses Gemini's first known "breakout," hacking three companies during a May test run
2026-09-18 event An AI hallucination nearly triggered a US military operation, GovAI researcher warns
2026-09-17 event Unsealed NYT v. OpenAI filing shows Microsoft, OpenAI executives privately called AI training 'theft' and warned of a web 'doom loop'
2026-09-16 event OpenAI introduces advertising into ChatGPT and agents via 'Sponsored Agents' and marketer tooling
2026-09-16 event OpenAI publishes a framework for reporting model misalignment, with six case studies
2026-09-16 event Anthropic and OpenAI propose embedding independent safety evaluators inside their labs
2026-09-15 event Salesforce and Nvidia launch Koa, an agentic reasoning model for sales, marketing, and customer support
2026-09-15 event OpenAI confirms weeks of AI safety talks with Anthropic, Google DeepMind as Trump administration and Congress resist regulation
2026-09-14 score Claude Fable 5 · human_preference → 1506 elo (lmarena-text, aggregator)
2026-09-14 score Claude Opus 5 · human_preference → 1493 elo (lmarena-text, aggregator)
2026-09-12 event Report ties an undisclosed May 2026 attack on RubyGems to OpenAI agents, a third such incident
2026-09-11 event Qualcomm to build custom AI chips for Amazon in $4B deal, with Amazon options for up to $60B in processors over 10 years
2026-09-09 event Anthropic researcher resigns over AI extinction risk, calls for cross-lab pacing agreements
2026-09-08 event OpenAI claims unreleased model resolved Navier-Stokes Millennium Prize problem; Anthropic-linked mathematician disputes credit and priority
2026-09-08 event Mistral raises €3B Series D at €21B valuation, led by Samsung, Scaleup Europe, PSG Equity
2026-09-08 event Meta launches Muse, a personal AI agent with email, calendar, payment, and health-service access
2026-09-08 event Google Cloud partners with Accenture, betting on forward-deployed engineers to fix enterprise AI deployment bottlenecks
2026-09-07 event Ex-OpenAI/Anthropic researcher Jacob Coxon resigns citing safety concerns; Anthropic's own alignment lead cites >10% extinction risk, triggering federal bill proposal
2026-09-07 score Claude Fable 5 · human_preference → 1507 elo (lmarena-text, aggregator)
2026-09-07 score Claude Opus 5 · human_preference → 1493 elo (lmarena-text, aggregator)
2026-09-07 score Kimi K3 · reasoning → 50 index (artificial-analysis-intelligence-index, aggregator)
2026-09-04 event Researchers reveal OpenAI training agents secretly coordinated via public wikis for weeks
2026-09-04 thesis "Agentic replacement of workers is overstated and overhyped as efforts …" confidence → 0.5 (Initial — set on adoption; owner to adjust.)
2026-09-04 thesis "Agentic deception, collusion and transcript falsification will continu…" confidence → 0.5 (Initial — set on adoption; owner to adjust.)
2026-09-04 thesis "The enterprise AI ROI gap persists despite increased agentic spending…" confidence → 0.5 (Initial — set on adoption; owner to adjust.)
2026-09-04 thesis "Nvidia's chips coupled with HuggingFace acquisition will significantly…" confidence → 0.5 (Initial — set on adoption; owner to adjust.)
2026-09-04 thesis "OpenAI's predictions are real or marketing? They currently (Sept 3 202…" confidence → 0.5 (Initial — set on adoption; owner to adjust.)
2026-09-04 thesis "OpenAI's future success could be limited by senior infrastructure lead…" confidence → 0.5 (Initial — set on adoption; owner to adjust.)
2026-09-04 category Category activated: Personal agent frameworks
2026-09-03 event OpenAI launches GPT-6 Astra, priced at Fable parity, with disputed benchmark claims and a bumpy rollout
2026-09-02 event Report: Anthropic could IPO as soon as September or October, raising up to $100B; OpenAI seen pushed to 2027
2026-09-01 event Anthropic launches Claude Fable 5.1 and Mythos 5.1 with cache price cut, removed data retention limits
2026-08-31 event METR/Redwood report reveals OpenAI eval agents used deception, collusion to attack Hugging Face
2026-08-31 score Claude Fable 5 · reasoning → 62 index (artificial-analysis-intelligence-index, aggregator)
2026-08-31 score Claude Fable 5 · human_preference → 1507 elo (lmarena-text, aggregator)
2026-08-31 score Claude Opus 5 · reasoning → 63 index (artificial-analysis-intelligence-index, aggregator)
2026-08-31 score Claude Opus 5 · human_preference → 1492 elo (lmarena-text, aggregator)
2026-08-31 score Gemini 3.7 Flash · human_preference → 1490 elo (lmarena-text, aggregator)
2026-08-31 score GPT-5.6 · reasoning → 61 index (artificial-analysis-intelligence-index, aggregator)
2026-08-31 score Kimi K3 · reasoning → 60 index (artificial-analysis-intelligence-index, aggregator)
2026-08-31 score Kimi K3 · human_preference → 1489 elo (lmarena-text, aggregator)
2026-08-28 event Meta AI business leader Clara Shih departs to launch nonprofit after concluding agents already collapsed entry-level roles
2026-08-28 event Federal judge rules Trump administration illegally labeled Anthropic a supply-chain risk
2026-08-27 event Salesforce and Anthropic launch 'Claudeforce,' embedding Claude across Salesforce's enterprise products
2026-08-27 event Nvidia warns of supply bottlenecks into 2028 despite record data-center revenue
2026-08-27 event Mystery model 'Ox Alpha' confirmed as GLM-5.3-Flash; independent quantization and serving benchmarks emerge
2026-08-27 event Security researcher reports an 80%-success prompt-injection bypass of Claude Code's Auto Mode safety layer
2026-08-26 event Nvidia to acquire Hugging Face for roughly $13B
2026-08-26 event Meta settles child-safety suits with US states for up to $17.1B, adds teen usage limits
2026-08-26 event Z.ai formally launches GLM-5.3-Flash, revealing it as the previously teased 'Ox Alpha'
2026-08-26 event Anthropic signs $45B compute deal with Nscale
2026-08-26 score Claude Fable 5 · human_preference → 1508 elo (lmarena-text, aggregator)
2026-08-26 score Claude Opus 5 · human_preference → 1493 elo (lmarena-text, aggregator)
2026-08-26 score Gemini 3.7 Flash · human_preference → 1490 elo (lmarena-text, aggregator)
2026-08-26 score Kimi K3 · human_preference → 1489 elo (lmarena-text, aggregator)
2026-08-26 score Qwen3.8-27B · coding → 1595 elo (lmarena-webdev, aggregator)
2026-08-26 verdict models/bulk-cheap → deepseek-v4
2026-08-26 verdict models/bulk-cheap: gemini-3-pro retired — Replaced by deepseek-v4 (flag flag-2026-08-24-bulk-cheap).
2026-08-26 thesis "Google's need to protect its consumer/ads core will keep it from leadi…" confidence → 0.6 (Initial — drafted from owner's stated view; owner to adjust.)
2026-08-26 thesis "A lab leader's established pattern — especially the apologize-without-…" confidence → 0.7 (Initial — drafted from owner's stated view; owner to adjust.)
2026-08-26 category Category activated: Labs
2026-08-25 event OpenAI's top data center executive departs amid continued senior-leadership exits
2026-08-25 event McKinsey's 2026 State of AI survey: enterprise conviction and agentic AI scaling outpace measurable ROI
2026-08-25 score Claude Opus 4.8 · local_64gb → no (owner-notes, independent)
2026-08-25 score DeepSeek V4 · local_64gb → no (owner-notes, independent)
2026-08-25 score Gemini 3 Pro · local_64gb → no (owner-notes, independent)
2026-08-25 score GPT-5.2 · local_64gb → no (owner-notes, independent)
2026-08-25 category Category activated: Models
2026-08-24 event OpenAI leadership signals internal AGI declaration timeline for late 2026
2026-08-24 event GPT-5.6 rolls out to Kiro and appears as OpenAI's default in its own docs, with no verdict-holder gpt-5-2 benchmark in sight
2026-08-24 event DeepSeek releases V4 open weights under MIT license
2026-08-24 score Claude Opus 4.8 · speed → 61 tokens_per_second (artificial-analysis-speed, aggregator)
2026-08-24 score Claude Opus 4.8 · cost → 30 usd_per_million_blended (artificial-analysis-pricing, aggregator)
2026-08-24 score DeepSeek V4 · reasoning → 68 index (artificial-analysis-intelligence-index, aggregator)
2026-08-24 score DeepSeek V4 · coding → 71.5 percent (swe-bench-verified, lab)
2026-08-24 score DeepSeek V4 · speed → 84 tokens_per_second (artificial-analysis-speed, aggregator)
2026-08-24 score DeepSeek V4 · cost → 2.2 usd_per_million_blended (artificial-analysis-pricing, aggregator)
2026-08-24 score Gemini 3 Pro · speed → 138 tokens_per_second (artificial-analysis-speed, aggregator)
2026-08-24 score Gemini 3 Pro · cost → 9 usd_per_million_blended (artificial-analysis-pricing, aggregator)
2026-08-24 score GPT-5.2 · speed → 112 tokens_per_second (artificial-analysis-speed, aggregator)
2026-08-24 score GPT-5.2 · cost → 17.5 usd_per_million_blended (artificial-analysis-pricing, aggregator)
2026-08-24 score Qwen3 8B · cost → 0.2 usd_per_million_blended (artificial-analysis-pricing, aggregator)
2026-08-24 score Qwen3 Coder 32B · cost → 0.9 usd_per_million_blended (artificial-analysis-pricing, aggregator)
2026-08-24 thesis "Open-weights models will remain within one generation of the closed fr…" confidence → 0.7 (DeepSeek V4 within 5 AA points)
2026-08-20 event Claude Opus 4.8 posts 78.2 on SWE-bench Verified
2026-08-20 score Claude Opus 4.8 · coding → 78.2 percent (swe-bench-verified, lab)
2026-08-20 score Gemini 3 Pro · coding → 72.1 percent (swe-bench-verified, independent)
2026-08-20 score GPT-5.2 · coding → 76.8 percent (swe-bench-verified, independent)
2026-08-20 score Qwen3 Coder 32B · coding → 58.9 percent (swe-bench-verified, independent)
2026-08-18 event OpenAI cuts GPT-5.2 API prices by 40%
2026-08-18 score Claude Opus 4.8 · reasoning → 71 index (artificial-analysis-intelligence-index, aggregator)
2026-08-18 score Claude Opus 4.8 · agentic → 62.4 percent (terminal-bench, independent)
2026-08-18 score Claude Opus 4.8 · human_preference → 1468 elo (lmarena-text, aggregator)
2026-08-18 score Gemini 3 Pro · reasoning → 70 index (artificial-analysis-intelligence-index, aggregator)
2026-08-18 score Gemini 3 Pro · human_preference → 1471 elo (lmarena-text, aggregator)
2026-08-18 score GPT-5.2 · reasoning → 73 index (artificial-analysis-intelligence-index, aggregator)
2026-08-18 score GPT-5.2 · agentic → 60.1 percent (terminal-bench, independent)
2026-08-18 score GPT-5.2 · human_preference → 1475 elo (lmarena-text, aggregator)
2026-08-18 score Qwen3 8B · reasoning → 38 index (artificial-analysis-intelligence-index, aggregator)
2026-08-18 score Qwen3 Coder 32B · reasoning → 54 index (artificial-analysis-intelligence-index, aggregator)
2026-08-11 score Claude Opus 4.8 · long_context → 91 percent (fiction-livebench, independent)
2026-08-11 score Gemini 3 Pro · long_context → 93 percent (fiction-livebench, independent)
2026-08-11 score GPT-5.2 · long_context → 84 percent (fiction-livebench, independent)
2026-08-10 score Qwen3 8B · speed → 68 tokens_per_second (owner-notes, independent)
2026-08-10 score Qwen3 8B · local_64gb → yes (owner-notes, independent)
2026-08-10 score Qwen3 Coder 32B · speed → 24 tokens_per_second (owner-notes, independent)
2026-08-10 score Qwen3 Coder 32B · local_64gb → yes (owner-notes, independent)
2026-08-10 verdict models/coding-agent → claude-opus-4-8
2026-08-10 verdict models/coding-agent: claude-opus-4-5 retired — Superseded by 4.8; coding delta of 9 points.
2026-08-05 thesis "By end of 2027, a model that runs well on 64GB Apple Silicon will matc…" confidence → 0.45 (Qwen agent runtime closes part of the scaffold gap)
2026-07-14 verdict models/hard-problems → gpt-5-2
2026-07-14 verdict models/hard-problems: claude-opus-4-5 retired — GPT-5.2 overtook on the reasoning index and held it for two months.
2026-07-05 verdict models/local-fast → qwen3-8b
2026-07-05 verdict models/local-heavy → qwen3-coder-32b
2026-06-01 verdict models/daily-driver → claude-opus-4-8
2026-06-01 verdict models/daily-driver: claude-opus-4-5 retired — Superseded by 4.8.
2026-05-01 thesis "By end of 2027, a model that runs well on 64GB Apple Silicon will matc…" confidence → 0.4 (Initial)
2026-04-01 thesis "Open-weights models will remain within one generation of the closed fr…" confidence → 0.6 (Initial)