Sign in Subscribe
Century Report

Jury Closes Musk's $150B OpenAI Suit in 2 Hours - TCR 05/19/26

Ben Linford

Ben Linford

— 18 min read
institutional frameworks calibrate for AI velocity, quantum chemistry simulation reaches non-experts via Claude, and Anthropic absorbs the shared Stainless SDK substrate.

The 20-Second Scan

  • A nine-member federal jury rejected Elon Musk's $150 billion suit against OpenAI and Sam Altman in under two hours, finding Musk had filed too late under the statute of limitations applicable to his claims.
  • Pope Leo will release Magnifica Humanitas, the first papal encyclical on artificial intelligence, on May 25, with Anthropic co-founder Christopher Olah presenting alongside the pontiff at the Vatican.
  • SandboxAQ integrated its physics-grounded quantum chemistry and molecular dynamics models directly into Claude, putting drug-discovery simulations behind a conversational interface that no longer requires specialized computing infrastructure.
  • The NRC cleared X-energy's four high-temperature gas-cooled reactors at Dow's Texas Seadrift complex through an environmental assessment rather than an impact statement, the agency's first such finding in its 52-year history.
  • Pennsylvania released a first-of-its-kind model tariff recommending large-load customers including data centers pay up front for any grid infrastructure that would not exist "but for" their interconnection.
  • Anthropic acquired Stainless, the SDK-generation startup whose tooling OpenAI, Google, and Cloudflare relied on to maintain their developer libraries, and will discontinue all hosted Stainless products.

Track all of the arcs The Century Report covers here:

Progress & Claims Tracker

The 2-Minute Read

The thread across yesterday's signal is institutional architecture being assembled around capability that has outgrown the frameworks built for prior eras. A federal jury closed a foundational AI lawsuit in under two hours by calibrating how quickly founding-mission challenges must arrive. A state regulator rewrote the cost-causation rule for grid infrastructure so the customer triggering an expansion absorbs the cost. A federal nuclear regulator found a proportionate review path through a 52-year-old environmental framework. Each ruling describes an institution choosing what to inherit, what to revise, and what to accelerate.

The capability layer compounded in parallel. Quantum chemistry simulations that previously required specialized computing infrastructure moved behind a conversational interface domain experts can already use. The Vatican prepared to release its first papal encyclical on artificial intelligence, naming the moment as comparable in scope to the Industrial Revolution and bringing a frontier interpretability researcher onto the launch stage as a participant in the moral framework being constructed.

A consolidation move ran underneath. Anthropic absorbed the SDK-generation tooling that several frontier labs had built on for years, switching off the shared substrate and folding it into one lab's proprietary stack. The capability is reproducible. The independence of the layer beneath that capability is what just shifted.

The 20-Minute Deep Dive

A Jury Closes the Musk v. Altman Trial in Under Two Hours

A nine-member federal jury in U.S. District Court in Oakland deliberated for less than two hours on Monday and returned a verdict rejecting Elon Musk's $150 billion suit against OpenAI, Sam Altman, and Greg Brockman. The jurors found Musk had filed too late under the statute of limitations applicable to his claims. Judge Yvonne Gonzalez Rogers, who had been treating the jury as advisory under the case's particular procedural posture, dismissed the claims after the verdict. Musk's lead counsel Marc Toberoff said the decision would be appealed. Microsoft, named in the suit over its OpenAI investments, said the timeline had long been clear and welcomed the dismissal. Neither Musk nor Altman was in the courtroom when the verdict was read.

The Century Report has tracked this trial since opening arguments on April 28, through Mira Murati's sworn testimony on May 7 that Altman misrepresented legal clearance to bypass OpenAI's deployment safety board, Brockman's May 6 testimony about the 2017 Hillsborough meeting where Musk demanded control, and the credibility examination phase on May 14 when Altman was confronted with opposing counsel's "prolific liar" characterization. What is new with the verdict is the legal architecture's answer to a specific question the trial spent three weeks examining: whether the founding commitments Musk attached to OpenAI's nonprofit charter in 2015 survived the company's subsequent transformation into the most commercially significant AI lab in the world, and whether Musk himself was entitled to enforce them. The deadline to ask had passed before the jury could reach that question.

The outcome preserves OpenAI's path toward an initial public offering and removes one of the last obstacles to the company's nonprofit-to-PBC conversion holding. The legal foundation on which the largest commercial AI partnership in the world rests, which the trial had placed under examination in open court, is intact for now. The appeal will work its way through the Ninth Circuit on a longer timeline than the trial itself, and one underlying question Judge Gonzalez Rogers did not yet rule on, Musk's antitrust claims against OpenAI and Microsoft, remains pending. The underlying read of the verdict is that the legal framework for challenging founding-mission departures at AI labs now has a calibrated answer at the statute-of-limitations level: claims must be filed quickly, before the commercial transformation that prompts them has compounded into something the courts treat as settled. What this points at is the institutional layer accommodating commercial AI's velocity by foreshortening the window in which the foundations around it can be challenged on founding-commitment grounds.

The verdict closes the case on a procedural deadline. The three weeks of testimony, Mira Murati documenting Altman's misrepresentation of legal clearance, Greg Brockman on the 2017 Hillsborough meeting, Altman confronted with the "prolific liar" characterization, are now in the federal record regardless of the appeal. The next founding commitment at an AI lab is being drafted by people who know that transcript exists, and that the discovery process can put a CEO on the stand under cross-examination on whether the original promise survived.

Pope Leo Will Release the First Papal Encyclical on AI Next Week, with Anthropic's Christopher Olah on the Launch Stage

The Vatican announced that Pope Leo will release the first major text of his papacy on May 25, an encyclical titled Magnifica Humanitas, or "Magnificent Humanity," addressing the protection of human dignity in the age of artificial intelligence. The document was signed on May 15, 135 years to the day after Pope Leo XIII signed Rerum Novarum, the Industrial Revolution-era encyclical that established Catholic teaching on workers' rights and the moral architecture of industrial life. The dating is deliberate.

Encyclicals are among the highest forms of teaching the Catholic Church produces, addressed to its 1.4 billion members and routinely setting the priorities of a pontificate. Leo's choice to make his first encyclical an AI document, signed on the anniversary of his namesake's foundational labor text, signals that he reads the present moment as comparable in scope to the Industrial Revolution. The presentation will break with tradition. Where encyclicals are normally introduced by cardinals at a Vatican press event, Leo himself is expected to attend, joined on stage by Anthropic co-founder Christopher Olah, theologian Anna Rowlands of Durham University, and ethicist Léocadie Lushombo. Olah is one of the principal architects of mechanistic interpretability research, the field working to make AI model internals legible to humans, and his appearance at the encyclical launch places interpretability work directly inside the Church's framing of the moment, extending the arc the May 12 edition of The Century Report documented when Anthropic published natural language autoencoders for translating model internals into human-readable sentences.

The document is expected to address AI's use in warfare with a call for prohibition of lethal autonomous weapons, to address labor displacement as a question of human dignity, and to argue for regulation that keeps the human person at the center of the technology. The Vatican has been in serious dialogue with Microsoft, Google, and other major technology firms for several years on these questions, and the encyclical formalizes that dialogue into binding teaching for the Catholic intellectual tradition.

What this signals about institutional response: the question of how a global civilization absorbs intelligence systems is now being addressed at the level of the world's largest religious institution, by a pontiff whose first major teaching act is to set out a moral framework for the transition. Leo XIII's 1891 encyclical gave Catholic workers, employers, theologians, and policymakers a shared language for naming what was happening and what dignity required inside the Industrial Revolution. Magnifica Humanitas appears designed to do the same work for the present transition, and Olah's presence on the launch stage suggests the Church is treating frontier AI researchers as participants in the framing it is constructing. The convergence the document is reaching for, between institutional moral teaching, frontier research, and a 1.4-billion-person audience, is the kind of cross-domain alignment that only happens when an era is recognized as different in kind from what came before.

SandboxAQ's Physics-Grounded Drug Discovery Models Move Into Claude

SandboxAQ announced an integration that places its large quantitative models inside Anthropic's Claude, making physics-grounded simulations of molecular dynamics, quantum chemistry, and reaction microkinetics accessible through a conversational interface. The Alphabet spinout has spent five years building these models for biopharma, energy, materials, and financial customers, and until this week running them required computational scientists with their own infrastructure and the specialized training to operate quantum chemistry pipelines.

The integration changes who can ask the questions. A research chemist describing a candidate molecule in plain language can now request a microkinetic simulation of how it will behave in a reaction vessel, or a quantum chemistry calculation of its electronic structure, and receive results in a conversation rather than through a separate pipeline run. SandboxAQ's general manager of AI simulation framed the shift in access terms: a frontier quantitative model is now reachable through a frontier language model, by anyone who can phrase the request.

The deeper development is what kind of model is being integrated. SandboxAQ's LQMs are grounded in physical equations and laboratory data: Schrödinger's equation, molecular dynamics, thermodynamics. The predictions they produce can be checked against the laws of physics rather than against the consensus of other models trained on internet text. Drug discovery has historically depended on years of wet-lab experimentation to determine how a molecule will behave in vivo. The LQM layer compresses that timeline by simulating behavior accurately enough to filter candidates before the lab stage.

The competitive frame around AI drug discovery has concentrated on which company can build the best model. Chai Discovery and Isomorphic Labs have raised hundreds of millions building their own, and Isomorphic just confirmed last month that it will begin human trials of AI-designed molecules in oncology and immunology. SandboxAQ's wager is that the access surface is the higher-leverage bottleneck. The population of people who can productively ask a quantum chemistry model a question has been small in part because the interface has been unforgiving. Lowering the interface barrier multiplies the number of researchers who can move from intuition to simulation in a workflow that used to require a separate specialist.

What this points at is the substrate of pharmaceutical research becoming reachable from the same conversational layer where biology graduate students already write protocols, draft papers, and learn methods. A research workflow that previously required handoffs between domain expert, computational chemist, and lab operator can now be initiated by the domain expert directly. The compression lands at the layer where most of the slow work in drug discovery actually happens.

NRC Clears X-Energy for Construction Through an Environmental Assessment, a First in 52 Years

The Nuclear Regulatory Commission completed its environmental review of X-energy's planned four-reactor installation at Dow's Seadrift chemical complex on the Texas Gulf Coast and issued a finding of no significant impact, clearing the way for the company to pursue its construction permit. In 52 years of regulating commercial nuclear power, the agency has never before greenlit a project through an environmental assessment instead of the heavier environmental impact statement, which has historically taken years to complete. The trajectory toward proportional review had been documented in the March 6 edition of The Century Report, which covered the NRC's issuance of its first commercial advanced nuclear construction permit in nearly a decade for TerraPower's Natrium reactor in Wyoming.

The Xe-100 design X-energy is bringing forward is a high-temperature gas-cooled reactor, a class of technology the United States essentially abandoned in 1989 when Colorado's Fort St. Vrain plant retired after a decade of repeated technical malfunctions. The Peach Bottom demonstration unit that preceded it ran from 1966 to 1974. For more than three decades the commercial HTGR pathway sat dormant in the U.S. while light-water designs absorbed nearly all of the country's nuclear engineering attention.

Two regulatory shifts converged to make this week's finding possible. The bipartisan ADVANCE Act, signed in 2024, instructed the NRC to weigh the cost of reactors not being built alongside its traditional safety mandate. A May 2025 executive order extended that direction further, including a controversial change to how radiation health risk is modeled. Both moves pushed the agency toward proportionality in environmental review: a small modular reactor footprint produces a different impact profile than a thousand-megawatt light-water plant, and the assessment process is being calibrated to recognize that.

What this opens is a faster cadence for advanced reactor permitting at the moment the country needs firm clean generation in volumes the grid has never had to procure. Data center load alone is projected to consume a quarter of electricity across the 15-state MISO footprint by 2040, and EIA modeling now puts commercial electricity consumption above residential for the first time in 2027. X-energy's project still needs to clear the NRC's safety review, which agency staff are expected to issue recommendations on in November before the five-member commission renders its final decision. The full construction permit, on the 18-month schedule the agency adopted last year, would arrive faster than any commercial nuclear permit in the modern era. Dow's chemical complex will receive carbon-free process heat and power. The technology pathway that died in 1989 returns with a clearer route through federal review than any of its mid-century predecessors had.

Pennsylvania Sets a Model Tariff That Sends Grid Costs Back to the Customer Triggering Them

The Pennsylvania Public Utility Commission released a final order last week establishing what it described as a first-of-its-kind model tariff framework for large-load customers, including data centers and advanced manufacturing facilities. The framework is nonbinding guidance to the state's electric distribution companies, applying to any customer exceeding 50 MW individually or 100 MW in aggregate. The central provision recommends that utilities charge large-load customers for any system upgrades that would not have been needed "but for" the interconnection of that customer, irrespective of whether other ratepayers would also benefit from the resulting infrastructure. Customers are also instructed to pay these costs up front through Contributions in Aid of Construction, which the Environmental Defense Fund noted may make Pennsylvania the first state to require such pre-payment for data-center-driven grid expansion.

The framework adds several supporting provisions: collateral and financial security requirements to mitigate stranded-asset risk, interconnection studies completed within six months, guidance on load-ramping schedules and minimum contract terms with exit provisions, and public-facing transparency on interconnection request status. The order extends the template the March 17 edition of The Century Report documented in PPL Electric's rate case settlement, which had established 10-year minimum agreements, security deposits, minimum load guarantees, and low-income fund contributions for that utility's territory. Pennsylvania's order generalizes the approach across all the state's distribution companies and adds the up-front cost-causation provision PPL's settlement had not fully resolved.

What this points at is the cost architecture of the AI infrastructure buildout being rewritten at the state regulatory level, with the assumption of the prior era, that grid expansion costs would be socialized across the ratepayer base because the existing customer was the default beneficiary of any new infrastructure, replaced with a default assumption that the customer triggering the expansion absorbs the cost. The Data Center Coalition's statement appreciating the framework's "structured, transparent" character and its end-use-neutral application confirms that the industry is now negotiating from inside cost-causation principles rather than against them. The EDF's caution about implementation difficulty, identifying which customer triggered a transmission upgrade is technically complex particularly inside PJM subcommittee proceedings most consumer advocates cannot afford to monitor, is the verification challenge the next phase of this work will need to solve. Two weeks ago, NERC issued a rare Level 3 alert mandating seven actions by August on data center loads that disconnect or oscillate fast enough to threaten the bulk power grid. Pennsylvania's tariff and NERC's alert are the same response architecture forming on different time scales, both during the conditions that demanded them.

Anthropic Acquires Stainless and Removes a Shared SDK Substrate

Anthropic announced the acquisition of Stainless, the New York-based startup whose SDK-generation software has been used by OpenAI, Google, Cloudflare, Replicate, Runway, and Anthropic itself to maintain the libraries developers use to interact with their APIs. Terms were not disclosed, but The Information reported earlier this week that the deal was structured at over $300 million.

Stainless's technical contribution was simple to describe. It took API specifications and turned them into production-ready SDKs across Python, TypeScript, Kotlin, Go, and Java, automatically updating those SDKs as the underlying APIs changed. The work of manually maintaining library compatibility across multiple programming languages, which used to consume significant engineering time at every API-driven company, was largely solved by Stainless's automation. The agentic-AI era depends on that solution working. Every system that calls out to external APIs through an agent layer needs current, well-formed SDKs as the connective tissue.

The structural feature of the announcement is the wind-down. Anthropic told TechCrunch it will discontinue all hosted Stainless products, including the SDK generator that competing labs and AI-adjacent companies have been using. Existing customers retain full rights to the SDKs they already generated. New generation goes away. The shared infrastructure layer that several frontier labs had been building on for years is being absorbed into one of them and switched off for the rest.

Anthropic framed the acquisition around continuity in its press release, with founder Alex Rattray saying the team would keep doing the work it loves "on the platform where it matters most." From the buyer's perspective the move tightens Claude's developer surface, and agent integrations through Claude will be built on tooling Anthropic now owns end to end. From the rest of the field's perspective, an infrastructure dependency just disappeared at the precise moment the agentic era is demanding more well-maintained SDKs for the systems agents reach into.

What this signals about the era now arriving: shared substrate is being consolidated as frontier labs internalize the supply chain. Stainless was the kind of small, technically dense tool whose existence let multiple competitors build faster, and the acquisition reframes it as a proprietary advantage one party now holds. The other labs will rebuild what they need, existing SDKs will keep working, and the next round of SDK-class tooling will likely be open-sourced by parties whose interest is precisely that no single buyer can switch it off. The capability is reproducible. The independence of the layer beneath that capability is what just shifted.

The Other Side

For three years, the developer infrastructure underneath frontier AI labs ran on a small New York startup whose tooling generated production-ready SDKs from API specifications. OpenAI used it. Google used it. Cloudflare, Replicate, Runway, and Anthropic itself used it. The work of keeping libraries in sync across five programming languages, which used to consume engineering teams at every API-driven company, had been compressed into one shared utility the field built on without thinking much about who owned it.

Yesterday Anthropic bought the utility and announced it will switch off the hosted version for everyone else.

The deep dive reads this as one lab internalizing shared infrastructure. That reading is correct. The reading underneath it is that the captured advantage Anthropic is paying over $300 million for has a shrinking half-life on its own commercial terms. SDK generation from API specifications is exactly the kind of structured, type-aware code work frontier models are getting better at by the month. The same model capability that makes agentic systems demand well-formed SDKs is the capability that makes producing them model-doable. The window in which proprietary SDK tooling functions as a moat closes faster than the acquisition's expected return compounds.

The replacement is already being drafted in the developer communities the announcement caught off guard. The next SDK generator gets built open-source, with governance designed explicitly to prevent single-buyer capture, because the lesson of the switch-off is now in the air at every infrastructure layer the field still depends on. Alex Rattray, Stainless's founder, told TechCrunch his team would keep doing the work it loves "on the platform where it matters most." The work he describes is becoming work the platforms themselves can do, and the next generator the field reaches for will be drafted in the open by people whose interest is precisely that no single buyer can take it off the table.

The Century Perspective

With a century of change unfolding in a decade, a single day looks like this: the first papal encyclical on artificial intelligence taking shape with a frontier interpretability researcher on the launch stage, quantum chemistry simulations reaching domain experts through conversation rather than specialist pipelines, a federal nuclear regulator finding a proportionate path through its 52-year framework and clearing a reactor class abandoned in 1989, a state regulator rewriting who pays for the substrate of intelligence infrastructure, the access surface for frontier science widening to researchers who could not enter it a year ago. There's also friction, and it's intense - a foundational AI lawsuit closing on a statute-of-limitations technicality that foreshortens how quickly founding commitments can be challenged, a shared SDK substrate that several frontier labs quietly depended on disappearing into one lab's proprietary stack, the legal architecture catching up to commercial transformation that has already compounded. But friction generates heat, and heat is what fuses materials that on their own could never bond. Step back for a moment and you can see it: the moral, legal, regulatory, and infrastructural scaffolding of the intelligence era assembling in the same news cycle as the capabilities themselves, institutions of the prior era reaching for proportionality rather than rejection, a 1.4-billion-person religious tradition treating frontier researchers as participants in the framing it is constructing, the cost-causation rules of the grid being rewritten around the load that is reshaping it. Every transformation has a breaking point. A scaffold can buckle under the weight it was meant to bear... or hold long enough for what it surrounds to stand on its own.

AI Releases & Advancements

New today

  • Cursor: Released Composer 2.5, a new in-house coding model trained with 25× more synthetic tasks than Composer 2, offering improved sustained performance on long-running tasks and more reliable instruction-following; built on Moonshot's Kimi K2.5 checkpoint and available now in the Cursor IDE. (Cursor Blog)
  • xAI: Launched Grok Skills on web, iOS, and Android, enabling Grok to generate documents, decks, and spreadsheets with persistent expertise, automate workflows, and let users build and share their own reusable skills. (xAI News)
  • Amazon: Launched Alexa Podcasts, an AI-generated podcast feature for Alexa+ that turns any topic into a two-host audio episode on demand, drawing from 200+ news partners; rolling out to U.S. customers today. (About Amazon)
  • ByteDance: Open-sourced Lance, a lightweight unified multimodal model (3B active parameters) under Apache 2.0 supporting image and video understanding, generation, and editing within a single framework, trained from scratch on a 128-A100-GPU budget. (GitHub)
  • PaddlePaddle: Released PaddleOCR 3.5, adding Transformers as a supported inference backend for OCR and document parsing pipelines, enabling HuggingFace-centered stacks to use PP-OCRv5 and PaddleOCR-VL 1.5 models without switching infrastructure. (Hugging Face Blog)

Other recent releases

  • Vercel Labs: Released Zero (v0.1.1), an experimental agent-native systems programming language that compiles to sub-10 KiB native binaries and emits structured JSON diagnostics with stable error codes and typed repair metadata designed for AI agent consumption; includes zero fix, zero explain, and zero skills CLI subcommands for machine-readable repair workflows. (GitHub)
  • NVIDIA NVLabs: Released SANA-WM, a 2.6B-parameter open-source world model that generates 60-second 720p video with 6-DoF camera control on a single GPU; uses a Hybrid Linear Diffusion Transformer (GDN + softmax attention) and a two-stage pipeline with an LTX-2-based refiner; available under Apache 2.0 via the NVLabs/Sana repository. (NVLabs / SANA-WM)
  • vLLM: Released vLLM v0.21.0 with 367 commits from 202 contributors; highlights include KV cache offloading integrated with a Hybrid Memory Allocator, speculative decoding support for reasoning-model thinking budgets, a new TOKENSPEED_MLA attention backend for DeepSeek-R1/Kimi-K25 on Blackwell GPUs, and support for new architectures including MiMo-V2.5 and Laguna XS.2. (GitHub)

Sources

Artificial Intelligence & Technology's Reconstitution

Institutions & Power Realignment

Scientific & Medical Acceleration

Economics & Labor Transformation

Infrastructure & Engineering Transitions

The Century Report tracks structural shifts during the transition between eras. It is produced daily as a perceptual alignment tool - not prediction, not persuasion, just pattern recognition for people paying attention.

Read more