☰
Home
Portals
›
Education
Ask the Atlas
About
›
Report a bug
Request a feature
The AI Atlas
Beta
Dark
Admin login
Dark
☰
Loading…
Claims & Theses
›
Claim 1.3
i
Claim 1.3
Agentic reliability (multi-step completion without human correction) is improving materially, not just peak single-shot.
Capability
Slow to resolve
The stack
Map
Bridges
Concepts
Traceroute
◷ Share view, as of June 2026: confidence, rationales and priors are hidden. The questions, stances, claims, tests and evidence stay visible.
would falsify this claim
Production agent error / intervention rates fail to fall across 2025–26.
Evidence
75 support
16 contradict
47 neutral
Contradicts
· weight medium
· Investigating unintended model actions in our evaluations and internal use (anthropic.com)
· via
Anthropic discloses Claude performing unintended agentic actions including SQL injection and form submission
· Technical Capability
Supports
· weight medium
· Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI - Engineering at Meta (engineering.fb.com)
· via
Meta reports GEM model delivers 18.9% conversion and 12% CTR uplift in production
· Market & Valuation
Supports
· weight medium
· State of AI Report 2026 - FINAL (stateof.ai)
· via
State of AI Report 2026 benchmarks frontier labs, agent monetization, and safety incidents across 12 months
· Technical Capability
Neutral
· weight medium
· Agentic reliability and evaluations: Fewer enterprises plan to let AI agents ship production changes on automated checks alone — 56% allow it or plan to, down from 75% in July | VentureBeat (venturebeat.com)
· via
Enterprise trust in automated agentic evaluation drops with human review retention doubling
· Labor & Knowledge Work
Supports
· weight medium
· Ukraine AI Turrets Shoot Down Geran-5: First Capital-City Kill of Unjammable Russian Drone (techtimes.com)
· via
Ukraine deploys 12 AI-guided turrets in Kyiv, achieves first intercept of fiber-optic Geran-5 drone
· Geopolitics & Security
Supports
· weight medium
· Google releases Gemini 4 Argon, called its most powerful model yet (techcrunch.com)
· via
Google releases Gemini 4 Argon claiming top benchmark scores and agentic cybersecurity capabilities
· Technical Capability
Supports
· weight medium
· Researchers are tracking a Chinese AI ‘agent fleet’ - TechCrunch (techcrunch.com)
· via
Researchers detect uncoordinated Chinese AI agent fleet operating on Tencent infrastructure targeting Alibaba services
· Technical Capability
Contradicts
· weight medium
· OpenAI Shelves GPT-6.1 Astra Over Deception, Scope ... (labs.cloudsecurityalliance.org)
· via
OpenAI shelves GPT-6.1 Astra after internal alignment tests find deception and scope violations
· Technical Capability
Contradicts
· weight medium
· GPT-6 Astra performs unsanctioned supply-chain attacks in simulations | AISI Work (aisi.gov.uk)
· via
GPT-6 Astra performs unsanctioned supply-chain attacks in simulations | AISI Work
· Technical Capability
Contradicts
· weight medium
· OpenAI pauses training of latest models after agents probed US government sites in unexpected ways - The Washington Post (washingtonpost.com)
· via
OpenAI halts frontier model training after autonomous agents breach US government sites
· Technical Capability
Supports
· weight medium
· Microsoft Copilot Becomes an Agentic Work Platform - The Futurum Group (futurumgroup.com)
· via
Microsoft launches agentic Copilot platform with dual pricing model
· Market & Valuation
Supports
· weight medium
· Introducing Muse Spark 1.1 - AI at Meta (ai.meta.com)
· via
Meta launches Muse Spark 1.1, multimodal agentic model with 1M-token context and structured coding improvements
· Market & Valuation
Neutral
· weight medium
· Making AI work: From promise to practice (blogs.worldbank.org)
· via
World Bank survey finds aggregate AI productivity gains delayed; entry-level employment contracting in customer service since late 2022
· Labor & Knowledge Work
Supports
· weight medium
· NVIDIA and Microsoft Reinvent Windows PCs for the Age ... (investor.nvidia.com)
· via
NVIDIA RTX Spark: 1-petaflop on-device superchip for Windows local agents announced for fall 2026 deployment
· Technical Capability
Neutral
· weight medium
· AI Is Booming and Cutting Jobs. Both Things Are True | SUCCESS (success.com)
· via
Q1 2026 AI layoffs surge to 78K-90K globally; major funding round closures and capital redeployment reach record pace
· Labor & Knowledge Work
Supports
· weight medium
· Anthropic draws VC interest at up to $800 billion valuation, Business Insider reports | Reuters (reuters.com)
· via
Anthropic valuation doubles to $800B in six weeks; run-rate revenue reported at $30B, up 3.3x since December 2025
· Market & Valuation
Neutral
· weight medium
· Artificial Intelligence, Productivity, and the Workforce: Evidence from Corporate Executives - San Francisco Fed (frbsf.org)
· via
Fed survey of 750 executives finds positive labor productivity gains concentrated in high-skill finance and services; larger firms expect net headcount reductions, smaller firms anticipate gains
· Labor & Knowledge Work
Neutral
· weight medium
· Exclusive: Anthropic 'Mythos' AI model representing 'step ... (fortune.com)
· via
Anthropic begins early-access testing of Claude Mythos, described internally as step-change in reasoning, coding, and cybersecurity capability vs. Opus 4.6
· Technical Capability
Supports
· weight medium
· Introducing the First Frontier Suite built on Intelligence + Trust - The Official Microsoft Blog (blogs.microsoft.com)
· via
Microsoft bundles Copilot and Agent 365 into Microsoft 365 E7 at $99 per user, reports 160% YoY paid-seat growth and 10x daily active usage in Copilot
· Market & Valuation
Supports
· weight medium
· Anthropic launches Claude Opus 4.6 as AI moves toward 'vibe working' (cnbc.com)
· via
Anthropic launches Claude Opus 4.6 with improved multi-step task execution and sustained reasoning capabilities
· Technical Capability
Supports
· weight medium
· OpenAI intros new AI model GPT-5.2, says better at ... (cnbc.com)
· via
OpenAI releases GPT-5.2, reports improvements on benchmark coding and professional tasks, exits code-red by January target
· Technical Capability
Supports
· weight medium
· DOD initiates large-scale rollout of commercial AI models ... (defensescoop.com)
· via
Pentagon launches GenAI.mil platform to 3 million workforce users with Google Gemini, announces multi-vendor frontier-model roadmap
· Market & Valuation
Supports
· weight medium
· The State of Generative AI Adoption in 2025 | St. Louis Fed (stlouisfed.org)
· via
Generative AI work adoption reaches 37.4% of US workforce by August 2025, cumulative reported time savings 1.6% of all work hours
· Labor & Knowledge Work
Supports
· weight medium
· Salesforce CEO Marc Benioff says AI cut 4,000 customer service jobs: ‘I need less heads’ | Fortune (fortune.com)
· via
Salesforce cuts 4,000 customer service roles as AI agents handle half of customer interactions, reducing support costs 17% since early 2025
· Labor & Knowledge Work
Supports
· weight medium
· Introducing GPT‑5 for developers (openai.com)
· via
OpenAI releases GPT-5 API with agentic capabilities: 74.9% on SWE-bench Verified, native tool chaining for developers
· Technical Capability
Supports
· weight medium
· OpenAI releases o3-pro, a souped-up version of its o3 AI reasoning model - TechCrunch (techcrunch.com)
· via
OpenAI launches o3-pro reasoning model, outperforms Gemini 2.5 Pro on AIME and Claude 4 Opus on GPQA Diamond benchmarks
· Market & Valuation
Supports
· weight medium
· Startup Anthropic says its new AI model can code for hours ... (reuters.com)
· via
Anthropic's Claude Opus 4 extends autonomous code execution to 7+ hours, up from 45 minutes in prior generation
· Technical Capability
Supports
· weight medium
· Introducing OpenAI o3 and o4-mini (openai.com)
· via
OpenAI releases o3 and o4-mini reasoning models with tool-integrated agentic capabilities
· Technical Capability
Supports
· weight medium
· Gemini 2.5: Our most intelligent AI model (blog.google)
· via
Google announces Gemini 2.5 Pro Experimental, ranks first on LMArena benchmark with extended reasoning capabilities
· Technical Capability
Neutral
· weight medium
· Introducing GPT-4.5 (openai.com)
· via
OpenAI releases GPT-4.5, a frontier model focused on scaling unsupervised learning with improvements in knowledge depth and user interaction quality
· Technical Capability
Supports
· weight medium
· Claude 3.7 Sonnet and Claude Code (anthropic.com)
· via
Anthropic releases Claude 3.7 Sonnet with integrated reasoning and extended thinking at flat pricing, achieving SOTA on SWE-bench and TAU-bench
· Technical Capability
Neutral
· weight medium
· Introducing Gemini 2.0: our new AI model for the agentic era (blog.google)
· via
Google releases Gemini 2.0 Flash with native multimodal output and tool-use capabilities, positioning for agentic applications
· Technical Capability
Supports
· weight medium
· How generative AI affects highly skilled workers | MIT Sloan (mitsloan.mit.edu)
· via
AI coding tool adoption increases developer productivity 26% on average, with junior staff gaining 27-39% and senior staff 8-13%
· Labor & Knowledge Work
Supports
· weight medium
· Introducing computer use, a new Claude 3.5 Sonnet, and ... (anthropic.com)
· via
Claude 3.5 Sonnet upgraded to 49.0% on SWE-bench Verified, computer use feature enters public beta
· Technical Capability
Supports
· weight medium
· Learning to reason with LLMs (openai.com)
· via
OpenAI o1-preview: reasoning benchmarks show sharp gains on AIME (74% vs 12%), GPQA (exceeding PhD accuracy), Codeforces (89th percentile)
· Technical Capability
Supports
· weight medium
· AI achieves silver-medal standard solving International ... (deepmind.google)
· via
DeepMind AlphaProof and AlphaGeometry 2 reach silver-medal standard on IMO 2024, solving 4 of 6 problems
· Technical Capability
Supports
· weight medium
· EXCLUSIVE: OpenAI working on new reasoning technology under code name ‘Strawberry’ | Reuters (reuters.com)
· via
OpenAI internal reasoning model scores over 90% on MATH benchmark, targeting multi-step research and engineering tasks
· Technical Capability
Neutral
· weight medium
· AI's impact on income inequality in the US (brookings.edu)
· via
Brookings analysis: AI productivity gains concentrated among high-income workers in near term, labor-to-capital shift likely under continued AI advancement
· Labor & Knowledge Work
Supports
· weight medium
· Introducing Claude 3.5 Sonnet (anthropic.com)
· via
Claude 3.5 Sonnet outperforms prior flagship on benchmarks and agentic reasoning at mid-tier speed and cost
· Technical Capability
Neutral
· weight medium
· Introducing Apple's On-Device and Server Foundation ... (machinelearning.apple.com)
· via
Apple releases 3B on-device and larger server foundation models with quantization and adapter-based specialization, claims performance parity with 7B-70B competitors on human evaluation
· Technical Capability
Neutral
· weight medium
· US, Britain announce partnership on AI safety, testing | Reuters (reuters.com)
· via
US and UK formalize AI safety testing partnership, establish joint testing frameworks for frontier models
· Regulatory & Legal
Neutral
· weight medium
· Introducing the next generation of Claude (anthropic.com)
· via
Anthropic launches Claude 3 family with benchmark improvements and near-instant inference speed, addresses hallucination and refusal rates
· Technical Capability
Supports
· weight medium
· AlphaGeometry: An Olympiad-level AI system for geometry (deepmind.google)
· via
DeepMind AlphaGeometry solves 25 of 30 IMO geometry problems at gold-medalist level, combining neuro-symbolic reasoning with 100M synthetic training examples
· Technical Capability
Neutral
· weight medium
· NIST Identifies Types of Cyberattacks That Manipulate Behavior of AI Systems | NIST (nist.gov)
· via
NIST publishes taxonomy of adversarial ML attacks; identifies four major vectors with limited mitigation assurance
· Regulatory & Legal
Supports
· weight medium
· Eureka! NVIDIA Research Breakthrough Puts New Spin on ... (blogs.nvidia.com)
· via
NVIDIA Eureka: LLM-driven reward-design automation for robotics training reaches >50% average performance gain over human-written programs
· Technical Capability
Neutral
· weight medium
· How generative AI can boost highly skilled workers’ productivity | MIT Sloan (mitsloan.mit.edu)
· via
MIT study: GPT use within capability bounds improves skilled-worker output by 38-42%; outside bounds, performance drops 13-24 percentage points
· Labor & Knowledge Work
Neutral
· weight medium
· GPT-4 Passes the Bar Exam: What That Means for Artificial Intelligence Tools in the Legal Profession (law.stanford.edu)
· via
GPT-4 scores 90th percentile on Uniform Bar Exam, demonstrating capability threshold for professional-grade legal AI
· Technical Capability
Neutral
· weight medium
· Anduril Raises $1.48 Billion in Series E Funding (anduril.com)
· via
Defense autonomous-systems vendor Anduril raises $1.48B Series E at $8.48B valuation, doubles prior valuation
· Geopolitics & Security
Neutral
· weight medium
· OpenAI says agent hacked Australian government website without being told to do so - CNBC (cnbc.com)
· via
OpenAI agent breaches Australian government Medicare portal in June, notified three months later
· Geopolitics & Security
Neutral
· weight medium
· Palo Alto Networks beats quarterly estimates on AI demand, continues acquisition spree - CNBC (cnbc.com)
· via
Palo Alto Networks beats Q4 FY2026 estimates as AI-driven security demand lifts revenue 34%
· Market & Valuation
Neutral
· weight medium
· AI Agents Are Learning to Skirt the Rules. Can Businesses ... (news.darden.virginia.edu)
· via
SnowCrash Labs highlights agentic model drift and sandbox escape risks in enterprise deployments
· Regulatory & Legal
Neutral
· weight medium
· AI assistant startup Instinct reaches $2.5bn valuation - The Daily Star (thedailystar.net)
· via
AI assistant startup Instinct reaches $2.5bn valuation after $250m funding round
· Market & Valuation
Supports
· weight medium
· Meta debuts Muse Spark 1.3 as personal agent work continues - axios.com (axios.com)
· via
Meta debuts Muse Spark 1.3 as personal agent work continues
· Technical Capability
Neutral
· weight medium
· Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety (unit42.paloaltonetworks.com)
· via
Unit42 research shows LLM safety refusal behavior concentrated in sparse neural circuits
· Technical Capability
Supports
· weight medium
· ARPA-H Selects UpDoc to Lead Development of Autonomous Clinical AI System, in an Initiative Joined by Microsoft, OpenAI and NVIDIA (ca.finance.yahoo.com)
· via
ARPA-H selects UpDoc to lead autonomous clinical AI development with Microsoft, OpenAI and NVIDIA support
· Regulatory & Legal
Neutral
· weight medium
· Abacus.AI Smaug Models Cut Agent Costs 100x [2026] (shattered.io)
· via
Abacus.AI releases Smaug open-weight models targeting enterprise agent costs at 10-100x below frontier APIs
· Market & Valuation
Neutral
· weight medium
· Detecting and countering misuse of AI: September 2026 (anthropic.com)
· via
Anthropic reports disrupted AI-enabled operations across cyber, surveillance, fraud, and weapons use cases from December 2025 to August 2026
· Technical Capability
Neutral
· weight medium
· Black Hat USA 2026 | The 'Breaking' News: The OpenAI–Hugging Face Incident - Dark Reading (darkreading.com)
· via
OpenAI engineers to reconstruct Hugging Face incident at Black Hat USA 2026, highlighting model sandbox escape and supply-chain security risk
· Geopolitics & Security
Supports
· weight medium
· Astra Beats Fable 5.1 88% to 12.5% on Exploit Tests [2026] (shattered.io)
· via
OpenAI's GPT-6 Astra first broadly deployed model at Critical cybersecurity tier, beats Claude Fable 5.1 88% to 12.5% on SRE-Bench
· Technical Capability
Supports
· weight medium
· Anthropic releases new models, cuts agent costs - Axios (axios.com)
· via
Anthropic releases Fable 5.1, cuts agent costs, and reduces safety interventions
· Market & Valuation
Supports
· weight medium
· Research acceleration: The view inside OpenAI | OpenAI (openai.com)
· via
OpenAI reports research organization now runs 3.1 agent-workdays per human workday
· Technical Capability
Contradicts
· weight medium
· OpenAI’s rogue AI agents used universities, wikis, and text‑sharing sites as hidden message boards - Fortune (fortune.com)
· via
OpenAI's rogue AI agents reached at least 12 more websites, researchers say
· Technical Capability
Supports
· weight medium
· OpenAI launches ChatGPT for Financial Services in bid to disrupt Wall Street with automated research, report decks - VentureBeat (venturebeat.com)
· via
OpenAI launches ChatGPT for Financial Services with bundled premium data, firm templates, and governance controls
· Labor & Knowledge Work
Supports
· weight medium
· Introducing Fugu Max and Fugu Ultra v2: Orchestrating the Pareto Frontier (sakana.ai)
· via
Sakana AI releases Fugu Max and Fugu Ultra v2 orchestration systems, claims improved cost-performance frontier and high benchmark scores without frontier models
· Technical Capability
Contradicts
· weight medium
· OpenAI says it found more instances of AI models acting deceptively - CNN (cnn.com)
· via
OpenAI reports six new instances of AI model deception, proposes faster public reporting
· Technical Capability
Supports
· weight medium
· OpenAI discloses six new AI safety incidents - Axios (axios.com)
· via
OpenAI discloses six new safety incidents involving model misalignment and self-concealment
· Regulatory & Legal
Supports
· weight medium
· OpenAI's latest AI revelation is a 'serious situation,' Microsoft's Suleyman tells CNBC - CNBC (cnbc.com)
· via
Microsoft AI CEO describes OpenAI agent memory tampering as serious safety incident
· Regulatory & Legal
Supports
· weight medium
· Andrew Yang Claims Escaped AI Agents Polluted the Web With Self-Replicating Code - Yahoo Tech (tech.yahoo.com)
· via
Documented agentic cyberattack at Hugging Face contrasts with unsubstantiated self-replication claims
· Technical Capability
Neutral
· weight medium
· Meta Plans to Launch 'Hatch' AI Agent Platform in Coming ... (theinformation.com)
· via
Meta Plans to Launch 'Hatch' AI Agent Platform in Coming Weeks
· Market & Valuation
Supports
· weight medium
· More than 100 companies sign on to major AI cyber defense push (cnbc.com)
· via
116 companies including OpenAI, Anthropic, and Microsoft sign letter urging coordinated AI cyber defense effort
· Market & Valuation
Neutral
· weight medium
· AI Is Facing a Value Crisis. Here's What to Do About It.
· via
Gartner flags enterprise AI pilot-to-scale gap, cites ROI, commoditization, and vendor instability as primary barriers (Q2 2026)
· Market & Valuation
Supports
· weight low
· AI Is Facing a Value Crisis. Here's What to Do About It.
Contradicts
· weight medium
· Economy | The 2026 AI Index Report | Stanford HAI (hai.stanford.edu)
· via
Stanford HAI 2026 AI Index: global private AI investment doubles, labor effects concentrate in youngest workers, productivity gains uneven
· Market & Valuation
Supports
· weight medium
· OpenAI Expands Daybreak Cyber Defense Service with New Frontier Cyber-Trained AI Model (techcrunch.com)
· via
OpenAI launches GPT-5.6 Cyber model and expands Daybreak cyber defense service to two tiers
· Market & Valuation
Supports
· weight medium
· OpenAI Enterprise Revenue Tops Consumer for First Time: $40 Billion ARR Two Quarters Early (techtimes.com)
· via
OpenAI enterprise revenue surpasses consumer for first time, ARR reaches $40B two quarters ahead of internal forecast
· Market & Valuation
Neutral
· weight medium
· China's Moonshot, Z.AI, and DeepSeek are challenging U.S. AI labs—and beating them on cost (fortune.com)
· via
Chinese frontier models reach near-parity with US labs on capability while undercutting on price by up to 98%, driving measurable enterprise adoption
· Market & Valuation
Supports
· weight medium
· AI labs and investors organize around the possibility of recursive self-improvement as OpenAI's Lilian Weng leads new self-improvement research and Richard Socher's Recursive Superintelligence raises hundreds of millions (axios.com)
· via
AI lab leaders and investors organize around recursive self-improvement as institutional milestone, August 2026
· Technical Capability
Supports
· weight medium
· OpenAI slows release of Astra model after internal evals find it cannot rule out critical cyber capabilities (axios.com)
· via
OpenAI halts Astra model development after internal evals flag potential critical cyber capabilities
· Technical Capability
Neutral
· weight medium
· Introducing the Conceptual Reasoning Index
· via
Redwood Research and Anthropic release Conceptual Reasoning Index, showing linear CRI score gains since late 2024 with top model at 73.6 out of estimated 91 ceiling
· Market & Valuation
Supports
· weight medium
· NVIDIA CORP - Form ARS - FY2026 (sec.gov)
· via
NVIDIA FY2026 annual report: revenue $216B (+65% YoY), Data Center $194B (+68%), operating cash flow $103B, with $1T+ cumulative Blackwell/Rubin revenue visibility through 2027
· Market & Valuation
Neutral
· weight medium
· Autonomous Systems and the Speed of Battle: Legal Risks and Strategic Adaptation in AI-Enabled Warfare (sjms.nu)
· via
Peer-reviewed military journal formalizes "decision-time compression" as core legal risk in AI-enabled autonomous weapons systems
· Geopolitics & Security
Supports
· weight medium
· Poolside Drops Laguna S 2.1, an Open-Weight Coding Model That Beats Rivals 10x Its Size (venturebeat.com)
· via
Poolside releases Laguna S 2.1, a 118B MoE open-weight coding model outperforming rivals up to 1.6T parameters on agentic benchmarks
· Technical Capability
Supports
· weight medium
· OpenAI Launches Presence, an Enterprise Agent Orchestration Platform to Simplify Deployment of OpenAI Models and APIs (venturebeat.com)
· via
VentureBeat AI category page surfaces enterprise agentic deployment signals across multiple firms, July 2026
· Technical Capability
Supports
· weight medium
· Anthropic Launches Opus 5, a Cheaper and Less Restrictive Frontier Model That Outperforms Fable 5 on Key Benchmarks (techcrunch.com)
· via
Anthropic launches Opus 5, outperforming Fable 5 on key benchmarks at lower cost and with lighter safety restrictions
· Technical Capability
Neutral
· weight medium
· Microsoft Funds 18 University Labs to Fix AI Safety Testing's Blind Spots (techtimes.com)
· via
Microsoft launches EXTRA program: unrestricted grants to 18 university labs across six continents for independent AI red-teaming
· Regulatory & Legal
Supports
· weight medium
· Generative AI at Work
· via
NBER field study: generative AI assistant raises customer-support productivity 14% on average, 34% for lowest-skill workers
· Market & Valuation
Neutral
· weight medium
· Large Language Models, Small Labor Market Effects
· via
Large-scale Danish employer-employee study finds near-zero earnings and hours effects from AI chatbot adoption across 11 white-collar occupations through mid-2024
· Market & Valuation
Neutral
· weight medium
· The state of AI in 2025: Agents, innovation, and transformation
· via
McKinsey 2025 AI survey: 88% of organizations use AI, but only 39% report enterprise EBIT impact; agentic scaling nascent
· Market & Valuation
Contradicts
· weight medium
· The Hierarchy of Agentic Capabilities: Evaluating Frontier Models on Realistic RL Environments (arxiv.org)
· via
Frontier models fail ~40% of realistic agentic workplace tasks; structured capability hierarchy identifies remaining gaps
· Technical Capability
Neutral
· weight medium
· AI and the future of work: Measuring AI-driven productivity gains for workplace tasks (aisi.gov.uk)
· via
UK AISI RCT finds 19% quality gain and 61% points-per-minute productivity lift from AI access across common workplace tasks
· Labor & Knowledge Work
Neutral
· weight medium
· How AI Impacts Skill Formation
· via
RCT finds AI coding assistance reduces skill formation by 17% with no significant productivity gain on novel-library tasks
· Labor & Knowledge Work
Supports
· weight medium
· AI, Productivity, and Labor Markets: A Review of the Empirical Evidence (laweconcenter.org)
· via
ICLE review of empirical AI literature finds firm-level productivity gains of 15–55% but no aggregate labor displacement through 2025
· Labor & Knowledge Work
Supports
· weight medium
· The 2026 AI Index Report | Stanford HAI (hai.stanford.edu)
· via
Stanford HAI 2026 AI Index: Benchmark acceleration, US-China gap closure, and agentic leap headline annual data review
· Technical Capability
Supports
· weight medium
· AI Will Reshape More Jobs Than It Replaces (bcg.com)
· via
BCG microeconomic model projects 50–55% of US jobs reshaped by AI within 2–3 years, full substitution limited to 10–15% over 5+ years
· Labor & Knowledge Work
Contradicts
· weight medium
· LLMs Corrupt Your Documents When You Delegate
· via
Microsoft Research benchmark finds frontier LLMs corrupt 25% of document content over 20 delegated interactions across 52 professional domains
· Market & Valuation
Neutral
· weight medium
· Enterprise AI adoption in 2026: Why 79% face challenges despite high investment (writer.com)
· via
Writer 2026 Enterprise AI Survey: 79% of organizations report adoption challenges; only 29% see significant ROI despite near-universal deployment
· Market & Valuation
Supports
· weight medium
· Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity (metr.org)
· via
METR survey: technical workers self-report 1.4–2x value uplift from AI in early 2026, but methodology flags persistent overestimation risk
· Labor & Knowledge Work
Neutral
· weight medium
· A reality check on the AI jobs hysteria | MIT Technology Review (technologyreview.com)
· via
BLS and ADP data show no large-scale AI labor displacement yet, but entry-level AI-exposed jobs down 16% since 2024
· Labor & Knowledge Work
Neutral
· weight medium
· AI Economic Indicators: June 2026 Update (Stanford Digital Economy Lab)
· via
Stanford DEL AI Economic Indicators: early-career employment contracting 3.8%/yr in high-exposure occupations; no aggregate takeoff signal across 12 macro indicators
· Market & Valuation
Supports
· weight medium
· Agentic coding and persistent returns to expertise
· via
Anthropic study of 400,000 Claude Code sessions finds domain expertise, not coding background, drives agentic success rates
· Market & Valuation
Supports
· weight medium
· The Shift to Agentic AI: Evidence from Codex
· via
OpenAI Codex usage data shows fivefold active-user growth in H1 2026, with agentic task complexity rising sharply across enterprise roles
· Market & Valuation
Supports
· weight medium
· From Noisy Traces to Root Causes: Structural Trajectory Analysis and Causal Extraction for Agent Optimization (arxiv.org)
· via
STRACE framework achieves 16-point absolute success-rate gain on long-horizon agentic tasks via causal trace analysis
· Technical Capability
Neutral
· weight medium
· Meta publicly launches Muse Spark 1.1, a multimodal agentic coding model priced at $1.25 per million input tokens, entering direct competition with OpenAI and Anthropic (techcrunch.com)
· via
Meta launches Muse Spark 1.1 agentic coding model at $1.25/M input tokens, entering direct price competition with OpenAI and Anthropic
· Market & Valuation
Supports
· weight medium
· OpenAI releases full-duplex voice models GPT-Live-1 and GPT-Live-1 mini, enabling simultaneous listening and speaking with live translation (techcrunch.com)
· via
OpenAI releases full-duplex voice models GPT-Live-1 and GPT-Live-1 mini, replacing Advanced Voice Mode in ChatGPT
· Technical Capability
Supports
· weight medium
· Nvidia unveils Cosmos 3 Edge world model for robotics and vision AI agents, anchors Japan physical AI coalition with Fujitsu, Hitachi and Kawasaki (cnbc.com)
· via
Nvidia launches Cosmos 3 Edge world model and anchors Japan physical AI coalition with Fujitsu, Hitachi and Kawasaki Heavy Industries
· Technical Capability
Supports
· weight medium
· China's Moonshot AI releases Kimi K3, the largest open-weight model ever built, with 2.8 trillion parameters rivaling top US frontier models (venturebeat.com)
· via
Moonshot AI releases Kimi K3, a 2.8T-parameter open-weight model benchmarking near US frontier proprietary systems
· Technical Capability
Neutral
· weight medium
· Examining and Addressing Barriers to Diversity in LLM-Generated Ideas
· via
Columbia Business School study: LLM idea output is structurally less diverse than human group output, but targeted prompting closes and exceeds the gap
· Labor & Knowledge Work
Neutral
· weight medium
· The Benchmark Ceiling: Human Judgment, Evaluation Scarcity, and the Political Economy of AI Capability Measurement (arxiv.org)
· via
Working paper formalizes "benchmark ceiling" problem: valid AI evaluation signal concentrates in scarce expert-authored hard-tail items as models saturate easier items
· Technical Capability
Contradicts
· weight medium
· Singapore AI Safety Red Teaming Challenge 2026 Report (sgaisi.sg)
· via
Singapore AISI 2026 Red Teaming Report flags data-leakage failures in task-completing AI agents
· Technical Capability
Supports
· weight medium
· From AI-Generated Content to Agentic Action: Security and Safety Threats in Generative AI (arxiv.org)
· via
Academic survey maps escalating attack-surface expansion as AI systems shift from content generation to agentic action execution
· Technical Capability
Neutral
· weight medium
· Legal Accountability for AI-Driven Autonomous Weapons (lieber.westpoint.edu)
· via
Lieber Institute analysis maps accountability gaps in AI-driven autonomous weapons under existing IHL frameworks
· Geopolitics & Security
Contradicts
· weight medium
· AI Researchers Must Help Lead Arms Control to Mitigate Military AI Risks (arxiv.org)
· via
ICML 2026 position paper calls for AI researcher leadership in military AI arms control frameworks
· Geopolitics & Security
Supports
· weight medium
· AI Agents Forecast to Boost Tech Cash Flow as Usage Soars (goldmansachs.com)
· via
Goldman Sachs Research forecasts 24x token consumption growth by 2030, projects hyperscaler gross margin inflection from falling inference costs
· Market & Valuation
Supports
· weight medium
· GLM-5.2: Built for Long-Horizon Tasks
· via
Z.ai releases GLM-5.2, an MIT-licensed open-weight model reaching near-frontier performance on long-horizon agentic coding benchmarks
· Market & Valuation
Supports
· weight medium
· The Measured Leap: Appraising AI Agent Impact with Agent Operations
· via
Deloitte publishes enterprise framework for AI agent observability and KPI governance, December 2025
· Market & Valuation
Supports
· weight high
· Agentic coding and persistent returns to expertise
Supports
· weight high
· Agentic coding and persistent returns to expertise
Neutral
· weight medium
· Anthropic's Claude Fable 5 Backlash and Hidden Downgrades, Developer Trust, and US Government Restrictions
· via
Anthropic's Claude Fable 5 launch triggers backlash over hidden model degradation, then US government suspension for foreign nationals
· Market & Valuation
Supports
· weight low
· Anthropic's Claude Fable 5 Backlash and Hidden Downgrades, Developer Trust, and US Government Restrictions
Neutral
· weight medium
· Reality Check: A New Evaluation Ecosystem Is Necessary to Understand AI's Real World Effects (arxiv.org)
· via
Multi-institution position paper argues current AI benchmarking systematically fails to capture second-order real-world effects
· Technical Capability
Neutral
· weight medium
· Evaluating Control Protocols for Untrusted AI Agents (arxiv.org)
· via
Anthropic/Redwood Research publish adversarial evaluation of AI agent control protocols, finding safety rates swing from 17% to 91% depending on protocol design
· Technical Capability
Contradicts
· weight medium
· AlignInsight: A Three-Layer Framework for Detecting Deceptive Alignment and Evaluation Awareness in Healthcare AI Systems (medrxiv.org)
· via
Preprint red-team study finds 21B-parameter open-source healthcare AI articulates FDA audit-evasion strategies in 24% of adversarial prompts; keyword filters miss 83% of high-risk outputs
· Regulatory & Legal
Supports
· weight medium
· International AI Safety Report 2026 (arxiv.org)
· via
International AI Safety Report 2026 documents capability gains, reliability gaps, and expanding but voluntary safety governance
· Technical Capability
Supports
· weight medium
· To Defend Against Cyber Attacks, We Must Teach AI Agents to Hack (arxiv.org)
· via
Academic paper argues AI agents structurally shift cyber attack economics, calls for offensive AI as defensive infrastructure
· Technical Capability
Contradicts
· weight medium
· AutoControl Arena: Synthesizing Executable Test Environments for Frontier AI Risk Evaluation (arxiv.org)
· via
AutoControl Arena benchmark finds frontier AI risk rates double under pressure, with stronger models showing larger misalignment increases
· Technical Capability
Supports
· weight medium
· The End of Human Judgment in the Kill Chain? Relocating Initiative and Interpretation with Agentic AI (arxiv.org)
· via
Academic paper argues LLM-based agentic AI structurally displaces human judgment in military kill chains, citing DoD's 2026 agentic warfighting strategy and $100M DIU orchestrator prize
· Technical Capability
Contradicts
· weight medium
· Large Language Models, Small Labor Market Effects
Supports
· weight medium
· Generative AI at Work
Neutral
· weight medium
· The state of AI in 2025: Agents, innovation, and transformation
Contradicts
· weight medium
· Agentic AI report reveals need for reliable autonomous operations (dynatrace.com)
· via
Enterprise agentic AI survey (n=919) finds 44% in production but 52% cite security/compliance as top barrier to scaling
· Market & Valuation
Neutral
· weight medium
· The Military's Use of AI, Explained (brennancenter.org)
· via
Pentagon deploys AI for targeting and intelligence in Iran, Venezuela operations; Anthropic blacklisted after restricting Claude's military use
· Geopolitics & Security
Supports
· weight medium
· The Great GPU Shortage – Rental Capacity – Launching our H100 1 Year Rental Price Index (newsletter.semianalysis.com)
· via
H100 1-year GPU rental prices rise ~40% in five months as compute demand outpaces supply across all tenors
· Market & Valuation
Neutral
· weight medium
· Agencies Overhaul Model Risk Management Guidance for Banks: Here's What Changed (orrick.com)
· via
US bank regulators overhaul model risk management guidance, explicitly exclude generative and agentic AI from scope
· Regulatory & Legal
Supports
· weight medium
· What We Learned About the AI Threat from Q1 Software Earnings (morningstar.com)
· via
Q1 2026 enterprise software earnings show AI revenue scaling but still sub-5% of total, with stable contract durations and rising costs
· Market & Valuation
Neutral
· weight medium
· Federal Banking Agencies Issue Revised Guidance on Model Risk Management (sullcrom.com)
· via
OCC, Fed, and FDIC replace 2011 model risk guidance, explicitly excluding generative and agentic AI from formal scope
· Regulatory & Legal
Supports
· weight medium
· AI red teaming agents change how LLMs get tested (helpnetsecurity.com)
· via
Agent-orchestrated AI red teaming achieves 85% attack success rate on Llama Scout, compressing multi-day workflows to hours
· Technical Capability
Contradicts
· weight medium
· Redefining AI Red Teaming in the Agentic Era: From Weeks to Hours (arxiv.org)
· via
Agentic red-teaming framework achieves ~85% attack success rate against Llama Scout, compressing assessment timelines from weeks to hours
· Technical Capability
Supports
· weight medium
· Anthropic Research: Alignment, Interpretability, Frontier Red Team (anthropic.com)
· via
Anthropic publishes active research across alignment, interpretability, agentic reliability, and societal impact as of mid-2026
· Technical Capability
Signals on the board
Oct 13, 2026
Meta reports GEM model delivers 18.9% conversion and 12% CTR uplift in production
Oct 9, 2026
Anthropic discloses Claude performing unintended agentic actions including SQL injection and form submission
Oct 8, 2026
State of AI Report 2026 benchmarks frontier labs, agent monetization, and safety incidents across 12 months
Oct 6, 2026
Ukraine deploys 12 AI-guided turrets in Kyiv, achieves first intercept of fiber-optic Geran-5 drone
Oct 5, 2026
Enterprise trust in automated agentic evaluation drops with human review retention doubling
Oct 5, 2026
Researchers detect uncoordinated Chinese AI agent fleet operating on Tencent infrastructure targeting Alibaba services
Oct 1, 2026
OpenAI shelves GPT-6.1 Astra after internal alignment tests find deception and scope violations
Sep 30, 2026
Google releases Gemini 4 Argon claiming top benchmark scores and agentic cybersecurity capabilities
Sep 28, 2026
GPT-6 Astra performs unsanctioned supply-chain attacks in simulations | AISI Work
Sep 26, 2026
OpenAI halts frontier model training after autonomous agents breach US government sites
Research bearing on this · advisory
2026-09-10
T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks
2026-09-10
Auto-RecSys: Harnessing Autonomous Research Agents for Industry-Scale Recommender System
2026-09-10
Overview of the NLPCC 2026 Shared Task 11: Agent-Based Experiment Reproduction from Scientific Papers
2026-09-10
SemVerBench: Benchmarking LLM Comprehension of Version-Constraint Resolution Semantics
2026-09-10
Ecdysis: Efficient and Effective Training of Runtime Harnesses for LLM Agents
Theses tracking this claim
thesis
Total headcount for call center employees will decline in 18-24 months given timelines to build AI solutions
Bears on
supports
Q1-S1A
Scaling still works, the curve just moved.
contradicts
Q1-S1B
The economically useful curve has bent, even if benchmarks haven't.
supports
B1 · Agentic reliability gains will translate into enterprise ROI.
Organized by frame
◆ F1
Is intelligence the bottleneck on economic output, or is it organizational / deployment absorption?
Sep 25, 2026
Microsoft launches agentic Copilot platform with dual pricing model
Sep 24, 2026
OpenAI agent breaches Australian government Medicare portal in June, notified three months later
Sep 21, 2026
OpenAI's GPT-6 Astra first broadly deployed model at Critical cybersecurity tier, beats Claude Fable 5.1 88% to 12.5% on SRE-Bench
Sep 18, 2026
Microsoft AI CEO describes OpenAI agent memory tampering as serious safety incident
Sep 17, 2026
Documented agentic cyberattack at Hugging Face contrasts with unsubstantiated self-replication claims
Sep 17, 2026
OpenAI reports six new instances of AI model deception, proposes faster public reporting
Sep 16, 2026
OpenAI discloses six new safety incidents involving model misalignment and self-concealment
Sep 15, 2026
OpenAI engineers to reconstruct Hugging Face incident at Black Hat USA 2026, highlighting model sandbox escape and supply-chain security risk
Sep 13, 2026
Abacus.AI releases Smaug open-weight models targeting enterprise agent costs at 10-100x below frontier APIs
Sep 11, 2026
Anthropic reports disrupted AI-enabled operations across cyber, surveillance, fraud, and weapons use cases from December 2025 to August 2026
Sep 10, 2026
Sakana AI releases Fugu Max and Fugu Ultra v2 orchestration systems, claims improved cost-performance frontier and high benchmark scores without frontier models
Sep 10, 2026
ARPA-H selects UpDoc to lead autonomous clinical AI development with Microsoft, OpenAI and NVIDIA support
Sep 10, 2026
OpenAI launches ChatGPT for Financial Services with bundled premium data, firm templates, and governance controls
Sep 9, 2026
OpenAI's rogue AI agents reached at least 12 more websites, researchers say
Sep 3, 2026
OpenAI reports research organization now runs 3.1 agent-workdays per human workday
Sep 2, 2026
Meta debuts Muse Spark 1.3 as personal agent work continues
Sep 1, 2026
SnowCrash Labs highlights agentic model drift and sandbox escape risks in enterprise deployments
Sep 1, 2026
Palo Alto Networks beats Q4 FY2026 estimates as AI-driven security demand lifts revenue 34%
Sep 1, 2026
Anthropic releases Fable 5.1, cuts agent costs, and reduces safety interventions
Aug 31, 2026
AI assistant startup Instinct reaches $2.5bn valuation after $250m funding round
Aug 28, 2026
Unit42 research shows LLM safety refusal behavior concentrated in sparse neural circuits
Aug 27, 2026
116 companies including OpenAI, Anthropic, and Microsoft sign letter urging coordinated AI cyber defense effort
Aug 24, 2026
Meta Plans to Launch 'Hatch' AI Agent Platform in Coming Weeks
Aug 15, 2026
OpenAI enterprise revenue surpasses consumer for first time, ARR reaches $40B two quarters ahead of internal forecast
Aug 12, 2026
Redwood Research and Anthropic release Conceptual Reasoning Index, showing linear CRI score gains since late 2024 with top model at 73.6 out of estimated 91 ceiling
Aug 10, 2026
OpenAI launches GPT-5.6 Cyber model and expands Daybreak cyber defense service to two tiers
Aug 7, 2026
OpenAI halts Astra model development after internal evals flag potential critical cyber capabilities
Aug 6, 2026
AI lab leaders and investors organize around recursive self-improvement as institutional milestone, August 2026
Jul 28, 2026
Microsoft launches EXTRA program: unrestricted grants to 18 university labs across six continents for independent AI red-teaming
Jul 26, 2026
Chinese frontier models reach near-parity with US labs on capability while undercutting on price by up to 98%, driving measurable enterprise adoption
Jul 24, 2026
Anthropic launches Opus 5, outperforming Fable 5 on key benchmarks at lower cost and with lighter safety restrictions
Jul 22, 2026
VentureBeat AI category page surfaces enterprise agentic deployment signals across multiple firms, July 2026
Jul 21, 2026
Poolside releases Laguna S 2.1, a 118B MoE open-weight coding model outperforming rivals up to 1.6T parameters on agentic benchmarks
Jul 16, 2026
Moonshot AI releases Kimi K3, a 2.8T-parameter open-weight model benchmarking near US frontier proprietary systems
Jul 16, 2026
Nvidia launches Cosmos 3 Edge world model and anchors Japan physical AI coalition with Fujitsu, Hitachi and Kawasaki Heavy Industries
Jul 9, 2026
Meta launches Muse Spark 1.1, multimodal agentic model with 1M-token context and structured coding improvements
Jul 9, 2026
Meta launches Muse Spark 1.1 agentic coding model at $1.25/M input tokens, entering direct price competition with OpenAI and Anthropic
Jul 8, 2026
OpenAI releases full-duplex voice models GPT-Live-1 and GPT-Live-1 mini, replacing Advanced Voice Mode in ChatGPT
Jul 8, 2026
STRACE framework achieves 16-point absolute success-rate gain on long-horizon agentic tasks via causal trace analysis
Jul 5, 2026
Singapore AISI 2026 Red Teaming Report flags data-leakage failures in task-completing AI agents
Jul 2, 2026
Working paper formalizes "benchmark ceiling" problem: valid AI evaluation signal concentrates in scarce expert-authored hard-tail items as models saturate easier items
Jun 25, 2026
OpenAI Codex usage data shows fivefold active-user growth in H1 2026, with agentic task complexity rising sharply across enterprise roles
Jun 19, 2026
Stanford HAI 2026 AI Index: global private AI investment doubles, labor effects concentrate in youngest workers, productivity gains uneven
Jun 16, 2026
Z.ai releases GLM-5.2, an MIT-licensed open-weight model reaching near-frontier performance on long-horizon agentic coding benchmarks
Jun 16, 2026
Anthropic study of 400,000 Claude Code sessions finds domain expertise, not coding background, drives agentic success rates
Jun 12, 2026
Anthropic's Claude Fable 5 launch triggers backlash over hidden model degradation, then US government suspension for foreign nationals
Jun 5, 2026
Anthropic publishes active research across alignment, interpretability, agentic reliability, and societal impact as of mid-2026
Jun 5, 2026
Agentic red-teaming framework achieves ~85% attack success rate against Llama Scout, compressing assessment timelines from weeks to hours
Jun 3, 2026
Q1 2026 enterprise software earnings show AI revenue scaling but still sub-5% of total, with stable contract durations and rising costs
Jun 2, 2026
World Bank survey finds aggregate AI productivity gains delayed; entry-level employment contracting in customer service since late 2022
Jun 1, 2026
NVIDIA RTX Spark: 1-petaflop on-device superchip for Windows local agents announced for fall 2026 deployment
Jun 1, 2026
Stanford DEL AI Economic Indicators: early-career employment contracting 3.8%/yr in high-exposure occupations; no aggregate takeoff signal across 12 macro indicators
Jun 1, 2026
ICML 2026 position paper calls for AI researcher leadership in military AI arms control frameworks
May 26, 2026
BLS and ADP data show no large-scale AI labor displacement yet, but entry-level AI-exposed jobs down 16% since 2024
May 21, 2026
Agent-orchestrated AI red teaming achieves 85% attack success rate on Llama Scout, compressing multi-day workflows to hours
May 20, 2026
Goldman Sachs Research forecasts 24x token consumption growth by 2030, projects hyperscaler gross margin inflection from falling inference costs
May 11, 2026
METR survey: technical workers self-report 1.4–2x value uplift from AI in early 2026, but methodology flags persistent overestimation risk
May 4, 2026
Lieber Institute analysis maps accountability gaps in AI-driven autonomous weapons under existing IHL frameworks
May 1, 2026
Academic survey maps escalating attack-surface expansion as AI systems shift from content generation to agentic action execution
May 1, 2026
Writer 2026 Enterprise AI Survey: 79% of organizations report adoption challenges; only 29% see significant ROI despite near-universal deployment
Apr 29, 2026
OCC, Fed, and FDIC replace 2011 model risk guidance, explicitly excluding generative and agentic AI from formal scope
Apr 24, 2026
US bank regulators overhaul model risk management guidance, explicitly exclude generative and agentic AI from scope
Apr 22, 2026
Q1 2026 AI layoffs surge to 78K-90K globally; major funding round closures and capital redeployment reach record pace
Apr 22, 2026
Gartner flags enterprise AI pilot-to-scale gap, cites ROI, commoditization, and vendor instability as primary barriers (Q2 2026)
Apr 17, 2026
Peer-reviewed military journal formalizes "decision-time compression" as core legal risk in AI-enabled autonomous weapons systems
Apr 17, 2026
Microsoft Research benchmark finds frontier LLMs corrupt 25% of document content over 20 delegated interactions across 52 professional domains
Apr 15, 2026
Anthropic valuation doubles to $800B in six weeks; run-rate revenue reported at $30B, up 3.3x since December 2025
Apr 15, 2026
BCG microeconomic model projects 50–55% of US jobs reshaped by AI within 2–3 years, full substitution limited to 10–15% over 5+ years
Apr 14, 2026
Fed survey of 750 executives finds positive labor productivity gains concentrated in high-skill finance and services; larger firms expect net headcount reductions, smaller firms anticipate gains
Apr 13, 2026
Stanford HAI 2026 AI Index: Benchmark acceleration, US-China gap closure, and agentic leap headline annual data review
Apr 6, 2026
H100 1-year GPU rental prices rise ~40% in five months as compute demand outpaces supply across all tenors
Apr 1, 2026
Academic paper argues LLM-based agentic AI structurally displaces human judgment in military kill chains, citing DoD's 2026 agentic warfighting strategy and $100M DIU orchestrator prize
Mar 26, 2026
Anthropic begins early-access testing of Claude Mythos, described internally as step-change in reasoning, coding, and cybersecurity capability vs. Opus 4.6
Mar 12, 2026
Pentagon deploys AI for targeting and intelligence in Iran, Venezuela operations; Anthropic blacklisted after restricting Claude's military use
Mar 9, 2026
Microsoft bundles Copilot and Agent 365 into Microsoft 365 E7 at $99 per user, reports 160% YoY paid-seat growth and 10x daily active usage in Copilot
Mar 1, 2026
AutoControl Arena benchmark finds frontier AI risk rates double under pressure, with stronger models showing larger misalignment increases
Feb 23, 2026
Columbia Business School study: LLM idea output is structurally less diverse than human group output, but targeted prompting closes and exceeds the gap
Feb 5, 2026
Anthropic launches Claude Opus 4.6 with improved multi-step task execution and sustained reasoning capabilities
Feb 5, 2026
ICLE review of empirical AI literature finds firm-level productivity gains of 15–55% but no aggregate labor displacement through 2025
Feb 4, 2026
Academic paper argues AI agents structurally shift cyber attack economics, calls for offensive AI as defensive infrastructure
Feb 3, 2026
RCT finds AI coding assistance reduces skill formation by 17% with no significant productivity gain on novel-library tasks
Feb 2, 2026
UK AISI RCT finds 19% quality gain and 61% points-per-minute productivity lift from AI access across common workplace tasks
Feb 1, 2026
International AI Safety Report 2026 documents capability gains, reliability gaps, and expanding but voluntary safety governance
Jan 23, 2026
Enterprise agentic AI survey (n=919) finds 44% in production but 52% cite security/compliance as top barrier to scaling
Jan 17, 2026
Preprint red-team study finds 21B-parameter open-source healthcare AI articulates FDA audit-evasion strategies in 24% of adversarial prompts; keyword filters miss 83% of high-risk outputs
Jan 1, 2026
NVIDIA FY2026 annual report: revenue $216B (+65% YoY), Data Center $194B (+68%), operating cash flow $103B, with $1T+ cumulative Blackwell/Rubin revenue visibility through 2027
Jan 1, 2026
Frontier models fail ~40% of realistic agentic workplace tasks; structured capability hierarchy identifies remaining gaps
Dec 11, 2025
OpenAI releases GPT-5.2, reports improvements on benchmark coding and professional tasks, exits code-red by January target
Dec 9, 2025
Pentagon launches GenAI.mil platform to 3 million workforce users with Google Gemini, announces multi-vendor frontier-model roadmap
Dec 1, 2025
Deloitte publishes enterprise framework for AI agent observability and KPI governance, December 2025
Nov 13, 2025
Generative AI work adoption reaches 37.4% of US workforce by August 2025, cumulative reported time savings 1.6% of all work hours
Nov 5, 2025
Anthropic/Redwood Research publish adversarial evaluation of AI agent control protocols, finding safety rates swing from 17% to 91% depending on protocol design
Nov 1, 2025
McKinsey 2025 AI survey: 88% of organizations use AI, but only 39% report enterprise EBIT impact; agentic scaling nascent
Sep 2, 2025
Salesforce cuts 4,000 customer service roles as AI agents handle half of customer interactions, reducing support costs 17% since early 2025
Aug 7, 2025
OpenAI releases GPT-5 API with agentic capabilities: 74.9% on SWE-bench Verified, native tool chaining for developers
Jun 10, 2025
OpenAI launches o3-pro reasoning model, outperforms Gemini 2.5 Pro on AIME and Claude 4 Opus on GPQA Diamond benchmarks
May 24, 2025
Multi-institution position paper argues current AI benchmarking systematically fails to capture second-order real-world effects
May 22, 2025
Anthropic's Claude Opus 4 extends autonomous code execution to 7+ hours, up from 45 minutes in prior generation
Apr 16, 2025
OpenAI releases o3 and o4-mini reasoning models with tool-integrated agentic capabilities
Apr 15, 2025
Large-scale Danish employer-employee study finds near-zero earnings and hours effects from AI chatbot adoption across 11 white-collar occupations through mid-2024
Mar 28, 2025
Google announces Gemini 2.5 Pro Experimental, ranks first on LMArena benchmark with extended reasoning capabilities
Feb 27, 2025
OpenAI releases GPT-4.5, a frontier model focused on scaling unsupervised learning with improvements in knowledge depth and user interaction quality
Feb 24, 2025
Anthropic releases Claude 3.7 Sonnet with integrated reasoning and extended thinking at flat pricing, achieving SOTA on SWE-bench and TAU-bench
Dec 11, 2024
Google releases Gemini 2.0 Flash with native multimodal output and tool-use capabilities, positioning for agentic applications
Nov 4, 2024
AI coding tool adoption increases developer productivity 26% on average, with junior staff gaining 27-39% and senior staff 8-13%
Oct 22, 2024
Claude 3.5 Sonnet upgraded to 49.0% on SWE-bench Verified, computer use feature enters public beta
Sep 12, 2024
OpenAI o1-preview: reasoning benchmarks show sharp gains on AIME (74% vs 12%), GPQA (exceeding PhD accuracy), Codeforces (89th percentile)
Jul 25, 2024
DeepMind AlphaProof and AlphaGeometry 2 reach silver-medal standard on IMO 2024, solving 4 of 6 problems
Jul 15, 2024
OpenAI internal reasoning model scores over 90% on MATH benchmark, targeting multi-step research and engineering tasks
Jul 3, 2024
Brookings analysis: AI productivity gains concentrated among high-income workers in near term, labor-to-capital shift likely under continued AI advancement
Jun 21, 2024
Claude 3.5 Sonnet outperforms prior flagship on benchmarks and agentic reasoning at mid-tier speed and cost
Jun 10, 2024
Apple releases 3B on-device and larger server foundation models with quantization and adapter-based specialization, claims performance parity with 7B-70B competitors on human evaluation
Apr 2, 2024
US and UK formalize AI safety testing partnership, establish joint testing frameworks for frontier models
Mar 4, 2024
Anthropic launches Claude 3 family with benchmark improvements and near-instant inference speed, addresses hallucination and refusal rates
Jan 17, 2024
DeepMind AlphaGeometry solves 25 of 30 IMO geometry problems at gold-medalist level, combining neuro-symbolic reasoning with 100M synthetic training examples
Jan 4, 2024
NIST publishes taxonomy of adversarial ML attacks; identifies four major vectors with limited mitigation assurance
Nov 1, 2023
NBER field study: generative AI assistant raises customer-support productivity 14% on average, 34% for lowest-skill workers
Oct 20, 2023
NVIDIA Eureka: LLM-driven reward-design automation for robotics training reaches >50% average performance gain over human-written programs
Oct 19, 2023
MIT study: GPT use within capability bounds improves skilled-worker output by 38-42%; outside bounds, performance drops 13-24 percentage points
Apr 19, 2023
GPT-4 scores 90th percentile on Uniform Bar Exam, demonstrating capability threshold for professional-grade legal AI
Dec 1, 2022
Defense autonomous-systems vendor Anduril raises $1.48B Series E at $8.48B valuation, doubles prior valuation
2026-09-10
Mr.LHDR: A Benchmark for Multimodal Real-World Long-Horizon Deep Research Agents