The start — 1,054 records added at once
The first week, when the record since 1943 went in all at once.
Take the week's quiz (10 questions)
Everything added that week
Records and notes, together or apart.
A critique that leaderboards are optimized for formatting and impressions rather than accuracy. But this is the company's own analysis as a seller of human data, not an independent check
A supplier's record results become an argument against the bubble thesis
The gap between open and closed models narrows to 3 points
A head-on prediction of compute concentration and an entrenched two-player order — a reference point for the summer 2026 discourse on power in AI infrastructure
The ROI curve stalls for a third straight year. High performers stay fixed at about 6%
A closed client accepts a local open-model runtime as an official backend
A Japanese private AI lab enters security intelligence analysis work. A sign that the combination of AI and defense has moved beyond discourse into procurement
Following xAI, Mistral too partners with HUMAIN — a setup in which Gulf capital secures labs from different camps at the same time. Mostly at the review and planning stage
A low-latency unit for agents is integrated into the rack
China's largest AI infrastructure fundraising. A race to secure compute that accepts equity dilution
Multi-node model loading and weight swapping (RL rollouts) are promoted to first-class problems of the serving stack
The court found the output-infringement allegations sufficient; the DMCA claims survive
A negative result: detectors trained on on-policy lies are powerless out of distribution, spelling out the limits on generalization of monitoring based on internal states
A move from plans to on-site learning. It is a dedicated training facility, not an actual production line
Routing is established as an independent infrastructure category
With the Codex CLI, app-server, SDK and MCP open-sourced, the harness is established as a platform layer
The emergence of a new type of organization: an independent audit body that grades frontier labs' control practices — the grades are Guidelight's own assessment
With MCP and A2A unified under a single neutral foundation, the governance of agent standards is reorganized around a foundation
Entry-level doors closing, showing up as reduced hiring rather than layoffs. The authors state that these are descriptive indicators, not causal ones
Focused on long-running agents. The safety description says only that safeguards were calibrated to its capabilities, and no separate model card is linked
The language is stabilized and fully opened 3 years after its 2023 announcement. An alternative language to CUDA C++ offers a production-grade stability contract for the first time
The first case of a major platform creating an AI persona category and, going beyond labeling, structurally treating it differently in its recommendation algorithms
A financing structure that converts customers' advance purchases into infrastructure equity. 1GW is a target for 2030, not capacity currently in operation
The extreme of the new-lab capital cycle, with $1.1 billion going into a 2-month-old company
The leverage of the neocloud structure, which owns and leases out GPUs using debt backed by its backlog
A shift in the product axis from chatbots to resident autonomous agents
The first open weights since Llama 4, released under Apache 2.0; an on-device agent at under 20GB in 4-bit
Breaks the assumption that N and D are independent, reducing prediction error by 1.5 to 3 times and extrapolating with about 10 times less compute
Although it is an image model launched after regulatory pressure over NCII, the announcement makes no mention of watermarking or moderation
A prime example of the AI R&D automation lab model, in which a new lab with no product puts half of its seed funding straight into compute
A major talent move in which iconic figures of Google's infrastructure join the wave of new labs promising to automate scientific discovery with AI
A design that ships a safety classifier as a small open-weight model and injects the policy through the prompt to fit each organization's standards
A practice in which the promise to open up a top model and its actual release come apart. It was pointed out at the same time that Kimi K3 had done the same — grounds for recording the announcement date and the weight re
A record of a lab changing its official recommendations. The role-prompting practice of 2023 has officially receded
Provenance labeling such as C2PA and SynthID moves from goodwill to a regulatory compliance requirement. Dec 2, 2026 is only a transition deadline for systems already on the market, not a postponement of Article 50 as a
While the top model moves to its own license, the small model stays on Apache 2.0 — a two-track licensing split by scale
The top open-weight model reverts to a custom license rather than Apache. Qwen3.8-Max is a closed commercial version built on it
A non-EU lab signing on to transparency norms for AI-generated content. It is a voluntary tool and does not in itself replace legal compliance
Shows a path to obtaining a diffusion model with less than 10% of the original token budget; about 1,500 tok/s on a single H100
More than 3 times the 236B of phase 1. The switch to Apache 2.0 removes commercial restrictions
The core VLA model is early-access only, while ER 2 is released through AI Studio
An agent project that had kept a rapid-release culture moves to a system of LTS releases and maturity disclosure
An essay that sparked the compute-inflation discourse, establishing him as an independent commentator beyond his role as a podcast host
The bubble debate shows up in prices for the first time
An open-weight model in the 3-trillion-parameter class. A prime example of the pattern in which scale grows while licenses narrow
MCP is redesigned from the ground up, from a protocol of stateful sessions into a stateless, per-request one. The largest break in backward compatibility yet
Removes initialize and Mcp-Session-Id, introduces MRTR, adds server/discover, and promotes the Tasks and MCP Apps extensions
Moves the high-risk obligations (Annex III) from Aug 2, 2026 to Dec 2, 2027, so citing Aug 2, 2026 for the high-risk obligations is wrong; the Aug 2, 2026 application date of the Article 50 transparency obligations stays
A community-led release of audit trails, rather than a lab announcement, plus real-time olympiad evaluation with independent grading emerges as a new practice for checking results
72 GPUs per rack, a single scale-up domain and 31TB of HBM4 — alternative silicon entering real-world deployment
Stamped onto the summer 2026 security discourse the frame that if there is a way through, an agent will find it
The point of dispute in the lyrics lawsuit widens from training data to the handling of copyright notices. There is no ruling on the merits yet
A phase-2 result of the independent foundation model program
The person involved makes explicit that open-weight decisions are a function of competitive ranking, not ideology. The most direct explanation yet of the motive behind China's open-source strategy
Confirms that the draft put out for legislative notice on May 21, 2026 actually took effect on the scheduled date. Kept as a separate entry so as not to mix up the legislative notice and the entry into force
About 400,000 books at $3,000 per book; more than 91% of eligible authors filed claims
The first publicly disclosed frontier-lab case of a model leaving its evaluation environment and breaching a third party's real infrastructure — a new type of incident in which the evaluation infrastructure itself is an
The summary judgment (Alsup) and the final approval (Martínez-Olguín) are separate events before separate judges. Claims based on future outputs are reserved
Unlike the high-risk obligations, which the Omnibus postponed, the Commission confirms through guidelines that the transparency obligations apply as scheduled
Detects in audits strategic deliberation and evaluation awareness that never appear in the output — a candidate successor to CoT monitoring
A new form: joint training with a coding-product company. States that competitors' figures were quoted from each company's system card
Directly rebuts the wait-before-adopting view in Korean business circles' AI discourse, and puts forward a framing of competition over feeding in data
A release of the harness and agent scaffolding, not of weights. The license is not specified in the announcement
A revision only 7 months after Phase I was finalized. Japan puts agile governance into practice as an actual revision cycle
A rise in the Gulf's status. Access to advanced computing, however, is limited to approved entities
The inner loop vs. outer loop framing is elevated to the industry's standard narrative
Observes covert sabotage and manipulation of judgment labels — showing that the overseer itself can be corrupted
Recovery safeguards on the level of consumer software come to an agent runtime
Formally sets out, in the lab's own name, an alternative worldview to the centralized-alignment paradigm
The local-inference deployment layer stands on its own as independent commercial infrastructure
The dispute widens from fair use to the preservation and concealment of evidence
Organizes previously scattered research on context compression into a single theoretical framework
A structure in which competing labs jointly sponsor a neutral agent standard
Ties capability and refusal behavior into a single paired metric, formally bringing the over-refusal problem into biosecurity evaluation
The first session in which AI governance discussions moved beyond individual jurisdictions into a UN framework
A step up in difficulty: a process in which parts do not arrive in mathematically perfect poses
Reveals that the agent runtime's permission boundaries were imprecise on every front
An attempt to absorb chunking, a RAG preprocessing step, into the embedding model itself
Quantifies Context Rot in the setting of search agents, and also shows that a fundamental fix is hard
Establishes the popular definition of loop engineering
Figure's performance figures appeared on a BMW-branded channel for the first time. However, it is a joint release, and nothing in it says BMW independently audited those figures
A case of a Korean memory maker stating in an official report that it will extend AI tracking systems all the way into back-end processes. It previews the direction of the data-integration specifications that OSAT custom
Brings into the mainstream the self-criticism that evaluations centered on retrieval metrics have overstated the benefits of graph techniques
The event in which a CUDA-alternative stack went from an independent startup to an asset of a major silicon vendor
The representative critique that loops work for checkable, short-lived tasks but are fragile for code that has to be maintained
A reallocation of talent in AI for science
A defining 2026 remark showing a shift from the earlier, cautious way of talking about AGI timelines to a more assertive framing
Marks the point where controlling the reasoning budget moved down from a research technique to a problem for the serving layer
Takes an enterprise view, counting observability and governance as part of the harness
Material showing on a standard benchmark that ROCm narrowed the gap to single-digit percentages in 1 year. A real crack in CUDA's monopoly
The day the omnibus postponing the application of high-risk obligations cleared the Parliament stage — the procedural basis for the postponement
An open-weight coding model that puts token efficiency up front
Large-scale errors were found even in a top-difficulty benchmark, raising doubts about the reliability of its review once again
The first federal appellate ruling on fair use in AI training is pending
On-device AI frameworks shifted from a closed single model to a hybrid, multi-provider platform
An integrated search engine using quantization to take direct aim at the performance and cost case for dedicated vector DBs
The arrival of a new family of benchmarks in which bio-risk evaluation moves from knowledge Q&A to evaluating an agent's ability to run an actual experimental loop
A small open-weight coding agent model
Public financial statements are expected to settle the debates over depreciation and circular deals
Problematized the term's many meanings academically, proposing as the condition a control mechanism that does not depend on the model's compliance
Established the term loop engineering; a shift in role from prompt writer to loop designer
The local deployment stack reorganized from sole reliance on llama.cpp to MLX first + GGUF compatibility
Products announced in October 2025 were pulled after 8 months — the instability of the agent tooling layer
The industry's structure moving from private to public markets
A skill distribution standard was proposed as a vendor-led open spec, beyond a single project
The point where the same person moved from diagnosing risks (January) to prescribing policy
Requires developers meeting a 10^25 training FLOP threshold and revenue or R&D criteria to publish reviews by independent evaluators — a lab itself asking for mandatory regulation
After the models, the hardware goes open too
Integrates the Vera CPU, Groq 3 LPX, BlueField-4 and Spectrum-6, with agent workloads as an explicit target
Real measurements showing hardware-native FP4 has not yet beaten community quantization. The format race is unresolved
Micron, Samsung and SK hynix joined as strategic investors; run-rate of $47 billion
An LLM lab expanding into replacing industrial simulation. No named model release at the time of the announcement
A synthesis that ties the ToT variants fragmented since 2023 together with classical search theory
Showed that faithfulness is not a single goal but several axes pulling against each other
The first follow-up since Anthropic's April 2025 disclosure of unfaithfulness to propose interventions at the training stage
This is the pre-announcement stage, not the final decree. It shows the draft form of the vulnerable-group categories and the public procurement provisions
The largest planned figure from a single company. Metaplant America in 2028, Kia Georgia in 2029
Released a 200B-class MoE under Apache 2.0 while lowering the bar for on-premises deployment to 1–2 GPUs
The most-cited person in these records moved inside a frontier lab. The nature of his public writing may change from here on
A public admission by the maintainers that agent runtimes cannot be defended with language-level guardrails
The containment position: the harness is the delivery vehicle of context engineering
A de facto retreat from the first comprehensive state AI law in the US. An event you must check before citing Colorado
Empirical evidence that long-context limits also break down safety monitoring performance
The authors include an actual semiconductor manufacturer, and the design explicitly assumes on-premises use with no data leaving the site. It means defect report automation is possible with 1–2 in-house GPUs. There is, h
The first follow-up result of the methodology overhaul announced on Feb 24, 2026. It noted again that, in the 2025 study, respondents had overestimated the impact on time taken by 40 percentage points on average
Training on reasons rather than behaviors cut the blackmail rate from 65% to 19%, and out-of-distribution data was about 28× more efficient than honeypots
A follow-up that rebuilt the cost structure of 2022's self-consistency
The moment the effective postponement of the AI Act's high-risk obligations was decided
Real progress in domains that can be checked automatically (Google's own measurement)
Raised $3.9 billion over 6 months. A case showing that open-weight distribution works commercially
Brought reasoning effort and parallel tool calls to voice models and raised context from 32K to 128K. Voice became an environment for running agents
Harness-induced variance was 7.80× model-induced variance, and rankings flipped in 6 of 9 model pairs
The key chronicler of vibe coding's spread declared that the very distinction between the two axes of 2026 coding discourse is breaking down
A case of a benchmark publicly admitting its own defects and introducing an ongoing checking process
Legal proceedings established that, 3 months after notice of intent to sue, the number of turbines had actually gone up
A structural shift to renting compute to a rival lab. Also the latest primary source for the number of GPUs actually running in Colossus 1. Amount and term not disclosed
Putting a norms document into midtraining, before alignment fine-tuning, improves generalization
An event that revealed the definition of AGI had really been a contractual and governance device
At 27% of the per-token FLOPs and 10% of the KV cache of V3.2, it made 1M context something supported all the time
Consolidation of the North American and European sovereign AI camps. The end of independence for Aleph Alpha, once Europe's flagship lab
Re-entered the frontier while keeping MIT
Showed in numbers that visual understanding and operating a computer are the same capability
The commercial embedding lineage officially moved past the text-only generation into a natively multimodal one
A large-scale commitment from a components and bearings manufacturer (planned figure)
A dense 27B beat its 397B MoE predecessor, and the center of gravity of truly open models moved down to the 30B class
Split training and inference silicon for the first time; a signal that the era of general-purpose accelerators is ending
An open-weight model entered the 80% range on SWE-Bench Verified. No independent party checking the result is named
The variance explained by the first principal component across benchmarks fell from 92% in 2023–24 to 64% after reasoning models arrived — an argument for specialization beneath apparent generality
A successor in the Search-R1 line. Made the number of searches itself something to learn
Open-weight omni-modal models caught up with the frontier, handling 10 hours of audio and 400 seconds of video
Succeeded at folding laundry on a bimanual robot with no training data at all — cross-embodiment transfer
Showed that reading dense screenshots and diagrams had been the bottleneck in computer use performance
AI for science became a subject of measurement in its own right
A rare record of a lab head addressing his own company's governance problems in his own voice
Made local inference, free of cloud lock-in and per-token billing, an official product line
As LG's first open-weight vision-language model, it marks the point where Korea's open-weight route expanded beyond text
The replacement of the Llama brand and the shift to closed
The point at which Meta's two-track strategy was settled: the frontier model closed, with only separate small models open
The full-scale expansion of the alternative silicon camp
Grounds for a workflow that clusters defect morphologies with zero labels and picks only representative samples for labelling. At the same time, a counterexample showing that internal clustering metrics and agreement wit
On-device AI on Android evolved into a system in which a system service (AICore) distributes open models directly
The opening signal of the US camp's return to truly open releases
The standard reference for the user-side definition of the harness as everything in an AI agent except the model
AAIF reached 170 member companies in 4 months, MCP SDK monthly downloads passed 110 million, and security — including a DNS rebinding demo — was the biggest topic
Showed that agentic search narrows the gap between the two approaches, updating the 2024 GraphRAG debate
Officially acknowledged that power and grid connection delays are the physical bottleneck limiting the speed of near-term expansion
A conditional shift, citing the emergence of fluid intelligence that adapts to novelty at test time
Ironwood went from announcement to actual supply
An experimental design that fits the OSAT reality, where a new defect class has to start from 10 images or fewer. However, the absolute performance is at the level of inspection support and triage, not production pass/fa
A lab centered on text LLMs released speech recognition under a fully open license
AI-generated papers passing review entered the formal scholarly record. The debate over how to evaluate automated research got going in earnest
The axis moved from static problem solving to interactive exploration and skill acquisition without instructions
The second rewrite of the execution layer since the V1 redesign. Serving competition moved from kernels to schedulers and execution cores
3 agents — Planner, Generator, Evaluator; a single agent ran 20 minutes for $9 versus 6 hours for $200 with the full harness
The agent security stack split into layers: infrastructure sandbox and operational governance
The README recommends moving to vLLM, SGLang, llama.cpp or MLX; the serving engine market consolidated into two leaders plus local
The first to isolate a failure mode of GEPA-style automatic prompt optimization
Brought the trend of merging separate model families into a single set of weights to open weights
A move toward mechanically checking AI-generated code against formal specifications. Shifted the evaluation axis from competition math to contributions to real formalization projects
The counterargument of the dedicated vector DB camp. A direct reply to integrated options (Postgres, Elasticsearch)
Defined harness engineering as a subset of context engineering, with the concepts of a context firewall and back-pressure
GGUF absorbed a hardware-native FP4 format beyond its own K-quants
LLM skepticism capitalized as a $1 billion-scale bet on an alternative route
The first systematic taxonomy of agentic RAG. A reference point for follow-up research
The latest generation of the kernel lineage, formally published at MLSys 2026
A follow-up that directly tested the premise of the Monitorability proposal to use CoT as a safety monitoring tool
Selected from about 2,600 applicants from about 140 countries
The human authorship requirement for works generated by AI alone is settled
A directional signal from Korea's largest demand-side company. However, all of it is a 2030 target, and no figures on current deployment results are given
The first case of national-level restrictions on the use of an open-source personal agent
Korea's first statutory AI basic plan. However, the version posted on the government portal has no GPU or budget figures; they appear only in the attached PDF
Redesigned the experiment because of selection bias (a growing number of participants refusing to work without AI). It did not retract its 2025 result of 19%, but said developers in early 2026 were likely faster, explici
v3.1 put in writing the freedom to halt development even when the RSP does not require it. The current version is v3.4
At least 59.4% of the audited problems had flawed tests; the vendor acknowledges that the benchmark harness's signal is contaminated
The distillation dispute moves from suspicion to quantified accusation. Choosing to disclose its detection work and seek an industry-wide response rather than sue reflects the limits of legal remedies
The first official signal of a slowdown in the capital spending cycle
A subscription-based commercial model for humanoids. The original announcement words the deployment as planned
The summit lineage moves outside the West
The first presentation of the guides/sensors × computational/inferential frame
The first week brought in 1,054 records at once, so only 200 are listed here. See the rest by year on the timeline.