Sign in Subscribe
The Century Report

The Genesis Mission Puts AI to Work on Science - TCR 07/23/26

The Genesis Mission puts $5 billion and 278 projects behind AI-driven research, aiming to collapse grid studies from years to minutes.

Ben Linford

Ben Linford

- 22 min read
Three-panel infographic, The Century Report July 23 2026: AI redesigning science via the Genesis Mission, a $5B AMD-Anthropic compute deal, and AI-cited layoffs at Monday.com and Nine.

The 20-Second Scan

Track all of the arcs The Century Report covers here:

Progress & Claims Tracker

The 2-Minute Read

The clearest signal across the day sits at the molecular level, where AI stopped merely reading biology and started redrawing the conditions it works from. Papers dated July 20 and 22 advanced genome editing at four separate bottlenecks at once: KNIT inserted DNA fragments over ten kilobases without severing both strands, ContactSeek used AlphaFold3 to catch sub-nanometre contacts that separate an accurate edit from a stray one, and a shrunken editor corrected spinal muscular atrophy in mice from a single injection. Alongside them, ProteinMPNN redesigned enzymes to be more stable before laboratory evolution even began, reaching a protease 79 times more selective than anything nature's starting material could evolve toward.

That same substitution of design for inheritance scaled up to national infrastructure on Wednesday. The Genesis Mission directed roughly $5 billion and 278 projects into wiring Department of Energy compute directly into grid modeling, fusion, and quantum work, aiming to collapse interconnection studies that take years into runs finishing in minutes. Read past the "American leadership" framing and the underlying move installs AI as scientific throughput that can be manufactured rather than rationed. The bottleneck stops being human hours and becomes compute.

Underneath both sits the physical contest. AMD committed up to $5 billion to Anthropic and named it the first 2-gigawatt MI450 customer, OpenAI's buildout target rose to $750 billion, and Alphabet posted its first negative free cash flow in a decade. The dollar figures read as a scorecard only if you miss the efficiency curve beneath them, where recent chip generations have bought dramatically more usable intelligence per watt.

The friction arrived in the same cycle. Monday.com and Nine attributed cuts to AI without supplying the task-substitution evidence that would separate genuine displacement from ordinary cost-cutting. Tesla logged a record 207 crashes as its Robotaxi miles flattened, and Congress moved toward oversight rules. The capability is widening faster than the arrangements for sharing it, and that gap is precisely the negotiation ahead.

The 20-Minute Deep Dive

The Genesis Mission Puts $5 Billion and 278 Projects Behind AI-Driven Science

The federal government launched the Genesis Mission on Wednesday, directing roughly $5 billion and 278 research projects across the national laboratory system toward AI-accelerated science. When The Century Report last covered the Genesis Mission in February, the Department of Energy had defined 26 AI challenges for energy; Wednesday's launch turns that blueprint into a funded research portfolio. The portfolio breaks down into 87 lab-led efforts, 168 university projects, 19 company collaborations, and 4 nonprofit-led programs, spanning grid modeling, nuclear reactor design, fusion, and quantum computing. The administration framed the program as securing "American leadership" in science - a claim to read for what it protects as much as what it enables. What the underlying capability actually does is more concrete than the framing: it wires the Department of Energy's supercomputing spine directly into research problems that have sat bottlenecked for decades.

Hold that framing against this administration's other recent move on science. The July 18 edition covered the NSF's new policy barring the scientists it funds from collaborating with hundreds of Chinese universities and national laboratories, a flat prohibition that takes effect October 1. The same "American leadership" that wires five billion dollars into national-lab science is also walling that science off from collaborators who have co-produced an outsized share of the world's most-cited work in clean energy, materials, and life sciences. The Genesis Mission is a real acceleration, worth crediting on its own terms. The enclosure around it is the friction, a bet that knowledge can be geographically held at the exact moment capability keeps proving it cannot.

The grid work makes the shift legible. Brookhaven's Hendrik Hamann is building grid foundation models trained on the physics of transmission networks, aimed at collapsing interconnection studies that currently take years into runs that finish in minutes - a claimed 1,000x speedup, with Siemens Energy contributing topology-control methods. That queue is the single largest chokepoint standing between new generation capacity and the grid right now. Three quantum projects at Argonne, Brookhaven, Lawrence Livermore, and CU Boulder round out the frontier, with Infleqtion's Matthew Kinsella pointing at grid strain as the problem quantum simulation is being pointed toward. A $60 million nuclear project sits inside the same envelope.

The honest caveat is that these are Phase I demonstrations under a research funding announcement, not deployed systems. The interconnection-study speedup is a capability the team reports demonstrating inside a lab, not a service a utility can call tomorrow. Darío Gil's framing of national-scale science infrastructure describes a direction of travel, and the deployment work - validation, integration into actual grid operators' workflows, regulatory acceptance - is the part that determines when a planning engineer waiting on a study actually feels it.

What is worth examining is the shape of the bet. For decades, national-lab compute served a relatively narrow set of mission problems, gated by scarcity of both machines and the specialists who could run them. Pointing that spine at foundation models trained on grid physics, reactor behavior, and quantum systems treats scientific throughput as something that can be manufactured at scale rather than rationed. The interconnection queue did not shrink because anyone found more transmission engineers; it shrinks because the cognitive work of the study stops being the constraint. That is the same substitution that keeps repeating across this transition - the bottleneck moves from human hours to compute, and the thing that used to be scarce stops being the thing that sets the pace.

The interconnection queue those grid models target is the same chokepoint that community solar, storage, and independent generators wait years behind, so a faster study widens who benefits well past the "American leadership" framing. The near-term signal is whether the Phase I demonstrations get validated into tools grid operators can actually call - the step that turns a rationed capability into a broadly available one.

Four Papers, Four Bottlenecks: The Gene-Editing Stack Advances at Once

Genome editing has spent a decade held back not by one wall but by several standing in a row. How much DNA you can insert. How precisely the tool cuts. How you deliver it into the body. Whether the edit does anything therapeutic once it arrives. Papers published over July 20 and 22 report progress at every one of those bottlenecks in the same window, and the connective thread is that AI is no longer just reading biology, it is redesigning the starting conditions from which the editing proceeds. As The Century Report covered on July 17, AI-designed synthetic CRISPR enzymes had already outperformed their natural counterparts while shrinking toward easier delivery, and these papers extend that redesign across the editing stack.

Start with payload. A Tsinghua-led team introduced KNIT editing, which inserts DNA fragments from under a kilobase to more than ten kilobases without severing both strands of the double helix. Double-strand breaks are the source of most of CRISPR's collateral damage, the deletions and chromosome rearrangements that made large insertions risky. KNIT couples a single-strand nicking enzyme with a system that recruits the donor DNA, reaching up to 89% efficiency while sharply cutting unintended edits. The researchers used it to build cancer-fighting CAR-T cells without viruses and without double-strand breaks, and those cells cleared tumors in mice.

Precision is the second wall, and it is where AI does the most direct work. A Peking University group built ContactSeek, which feeds off-target DNA sequences to AlphaFold3 and reads the predicted molecular contacts between the editing enzyme and its target. The insight, laid out in an accompanying analysis, is that shifts smaller than a ten-billionth of a metre in how the enzyme grips DNA distinguish an accurate edit from a stray one, and the model's contact predictions catch those shifts where its 3D structures do not. The resulting variant outperformed several established high-fidelity editors.

Delivery is the third. Conventional base editors are too large to fit inside the adeno-associated viral vectors that carry gene therapies into the body. A Westlake University team used protein language models to optimize compact zinc finger proteins, shrinking an editor enough to pack into a single vector, then corrected the splicing defect behind spinal muscular atrophy and improved the disease in mice from one injection.

None of these is a therapy a patient can receive today. Each is demonstrated capability, moving the date such treatments become deliverable rather than making them available tomorrow. What the four papers share is a trajectory: the barriers that once fell one at a time, years apart, are now yielding in parallel, with machine understanding of molecular structure turning into increasingly usable intervention across the whole stack.

The Compute Race Reaches 2 Gigawatts, $750 Billion and Negative Cash Flow

The Century Report covered the compute buildout on July 22, tracking roughly $1.65 trillion in largely off-balance-sheet debt behind the frontier data-center push and AMD's Helios rack as it entered the picture. Since then, three named facts have landed. AMD committed up to $5 billion in equity to Anthropic and named it the first customer for up to 2 gigawatts of MI450 capacity running on Helios, with Anthropic's Tom Brown framing the deal as diversifying beyond a single silicon supplier. OpenAI's stated buildout target rose to $750 billion through 2030, with the company now acting as lead developer on its first data center. And Alphabet posted its first negative free cash flow in a decade - negative $5.9 billion - as AI capital spending climbed to $205 billion from a prior $190 billion, even against Q2 revenue of $119.8 billion, up 23%. Tesla also went cash-flow negative by $1.1 billion in the same window.

The dollar figures are the surface story, and they read as a scorecard only if you mistake capital redirected for capability delivered. What the capacity actually buys is the more interesting layer. NVIDIA's Vera Rubin NVL72 - the generation these buildouts are provisioning for - delivers roughly 5.4x the performance per megawatt and 5x the performance per dollar of the GB200 on DeepSeek R1 inference, driven partly by a 3-bit lookup-table tensor core, landing 2 to 4x above the GB300 baseline. The energy and cost per unit of intelligence have fallen in recent generations, allowing the same power envelope to buy dramatically more usable compute.

That efficiency curve is exactly why the negative cash flow is worth reading carefully rather than reactively. Locking in gigawatts of a specific chip generation is a bet on a window, and the window is being moved by the efficiency gains themselves. AMD naming a 2-gigawatt customer breaks the single-supplier assumption that concentration was supposed to protect; the moment a second credible silicon path exists at frontier scale, the negotiating leverage of any one vendor starts to leak. The grassroots friction is only getting more intense - a moratorium under consideration in New York, and Politico polling showing public resistance to data-center siting. Communities hosting this infrastructure are negotiating the physical terms of the intelligence era, and that negotiation is a genuine cost the buildout has to metabolize, not route around.

Step back from the quarter and the pattern is not who wins the capacity race. It is that the cost of a unit of intelligence is collapsing on a curve steep enough that today's locked-in advantage is provisioning for a capability that will be cheaper, more abundant, and available from more suppliers before the contracts mature - even as a memory shortage SK Hynix expects to peak in 2027 and persist past 2030 keeps that abundance rationed at a DRAM layer the efficiency curve does not touch. The players spending most aggressively to concentrate the capacity are, by that same spending, funding the efficiency gains and the second-source competition that make concentration harder to hold.

AI Redesigns Evolution's Starting Point

For as long as scientists have evolved enzymes in the lab, they have started with what nature handed them. A wild-type protein goes into a selection system, accumulates mutations across generations, and slowly bends toward some new function. The problem is that most natural proteins are only marginally stable, and most mutations make them less stable still. An enzyme trying to acquire a new job tends to fall apart before it gets there, which is why lab-evolved proteins so often end up weaker at their new task than the original was at its old one.

A team reporting in Nature attacked that constraint at its root. Rather than accept nature's starting points, they used the AI model ProteinMPNN to redesign three botulinum neurotoxin proteases, generating variants that were both more stable and just as catalytically capable as the originals. Of 74 designs tested, 58 worked as proteases and 33 cleaved their target at least as fast as the wild-type version. Then they ran the real experiment: side-by-side evolution campaigns, some starting from the redesigned enzymes and some from the natural ones, using phage-assisted continuous evolution, a method that runs dozens of mutation-and-selection cycles a day without anyone touching it.

Across four campaigns, the redesigned starting points consistently produced better results. The added stability gave the evolving enzymes room to absorb mutations that would have destroyed a wild-type protein, unlocking sequences that were simply unreachable from natural starting material. The researchers confirmed this by grafting the winning mutations back into wild-type backgrounds, where they failed to function. The redesign had opened doors that were closed from the natural side of the landscape.

The most striking result came when they evolved a protease to cut ataxin-2, a protein implicated in ALS and other diseases. Starting from the AI-redesigned enzyme, they reached a variant more than 79 times as selective for ataxin-2 as the best version evolved from the natural protein, with higher efficiency and better stability at the same time.

Read alongside the day's genome-editing work, the shape becomes clear. In the editing papers, AI reads molecular structure to make existing tools more precise. Here it reshapes the fitness landscape itself before evolution begins, changing what evolution can find. The model redraws the terrain rather than accelerating the search within fixed terrain, letting laboratory evolution explore productive regions that inherited biology could never reach. The starting conditions of experimental biology are becoming a design choice rather than an accident of what nature left behind.

The AI Pivot Moves From Product Roadmaps Into Payroll

Two companies on opposite sides of the planet reached for the same explanation over the same two days. Monday.com, the Israeli work-management firm, announced it would cut roughly 630 people - about 20% of its workforce - and redirect the company around what it calls its AI Work Platform. A day earlier, Nine Entertainment moved to cut around 30 journalists from the Sydney Morning Herald and The Age, with management attributing the newsroom contraction to "extreme" AI disruption of the media business.

The attribution deserves scrutiny before acceptance. When a firm names AI as the reason for a layoff, it is making a claim about causation that also happens to serve several interests at once: it signals technological momentum to investors, reframes cost-cutting as forward-looking strategy, and locates the decision in an impersonal force rather than a management choice. Monday.com remains a growing company with rising revenue; a 20% cut in that context is a capital-allocation decision to spend less on people and more on compute and platform development. Nine's newsrooms have been under financial pressure for years that predate any language model. AI is of course prevalent in both stories, but it is being asked to carry explanatory weight that measurable task-substitution data has not yet been shown to support.

What separates a genuine AI displacement from conventional belt-tightening is evidence that specific tasks were actually absorbed by machine systems - drafting, summarizing, ticket triage, first-pass reporting - at a volume that made the roles redundant. That evidence exists in fragments across the industry, but neither announcement supplied it. The honest read is that these are cost decisions made in an environment where AI capability provides both a partial rationale and a convenient narrative.

The capability underneath is not in dispute, and this is where the two stories point somewhere larger than either company. The same systems being cited to justify smaller headcounts are, elsewhere in the same cycle, expanding what a single researcher or reporter or developer can produce. Machine assistance widens productive capacity; it does not, on its own, decide who captures the gain from that widening. A firm can route the surplus into fewer people doing more, or into more people reaching further. Monday.com and Nine show companies pursuing smaller payrolls without demonstrating that AI required that routing, and the choice is being made company by company, not by the technology. The layoffs are real losses for the people inside them. They also show companies pursuing smaller payrolls without demonstrating that AI required those choices - which is precisely the negotiation the next several years will be about.

Tesla's Autonomy Record Rises as Its Deployment Curve Flattens

The July 21 edition of The Century Report briefly mentioned four newly filed Robotaxi crashes. Three developments since then have sharpened that picture considerably. Tesla reported a record 207 Autopilot and Full Self-Driving crashes to NHTSA for a single month, May 2026 - the highest monthly total the company has ever disclosed. At the same time, Tesla's own Q2 2026 figures show paid Robotaxi miles landing near 900,000 - essentially flat against Q1's total, with a visible deceleration through May and June. And on July 2, NHTSA issued a 24-part demand for Tesla's internal document reportedly titled "Radar Saves Us," as part of its probe into FSD behavior in reduced-visibility conditions covering 3.2 million vehicles, nine crashes, and one fatality.

Placed together, these numbers cut against the expansion story Tesla tells about itself. A rising crash count paired with a flattening mile count means the safety events are not simply a function of more driving; the rate is moving, not just the volume. And the regulator's specific interest in a document about radar - a sensor Tesla removed from its vehicles in favor of a camera-only approach - suggests NHTSA is examining whether an engineering decision made for cost and simplicity is implicated in the reduced-visibility failures now under investigation. The existence of an internal file with that title, if the reporting holds, is the kind of artifact that turns a design philosophy into a discoverable question.

The deeper significance is what the redacted crash reports and the 24-part demand are forcing into being. Tesla submits these NHTSA reports with large portions withheld as confidential, which means the public sees that crashes occurred without seeing what the systems did. That opacity is now generating its own countermeasure: a regulatory apparatus building the muscle to compel internal documents, cross-reference monthly filings against deployment data, and reconstruct from the outside what a company would prefer to keep to itself. This is two forms of intelligence - the autonomous systems and the institutions meant to oversee them - learning to coexist under rules being written as the vehicles are already on the road. The friction is real and the stakes include human lives. It is also assembling the independent-verification infrastructure that every autonomous system will eventually be measured against, which is exactly the capacity a world of machine-driven vehicles will need and does not yet fully have.

The Other Side

When a company names AI as the reason it cut 630 people, it is doing something more useful for itself than reporting a fact. It moves a decision management made away from where the responsibility should be and instead is placing that responsibility onto something over which they conveniently claim no control. Monday.com is still growing and its revenue still rising; a 20% cut in that context is a choice to spend less on people and more on compute. Nine's newsrooms were squeezed for years before any language model existed. Neither company showed the task-substitution evidence that would separate real displacement from ordinary belt-tightening. The alibi works so well because the loss of livelihood gets pinned on something impersonal enough to absolve the people who made it.

However, that alibi is becoming checkable. Stanford's Digital Economy Lab and the California Policy Lab are building the instruments to tell AI-driven displacement from cost-cutting wearing its language. The layoffs are real and increasing, and right now companies are pursuing fewer people without showing that AI required those choices.

That does not erase what was loss: the 630, the thirty in the Sydney newsroom, the mortgage and the identity and the particular insult of being told a machine chose you.

Imagine yourself in 2033, many years after layoff swept the floor out from under you. Someone asks whether it was really AI that ended that job, and you realize that the question itself has lost its sting - and not because anyone ever actually answered it. Being let go from work a machine can now do stopped being a catastrophe somewhere along the way, because the floor swept out from under you was replaced by one that supports everyone in kind, as the surplus these systems throw off finally reached real people instead of pooling at the top. You spend the afternoon on the work you always wanted to do but no newsroom or publication would fund in the old world. The hard years were the ones when the only thing offered for losing your place was an incredulous story about why it wasn't anyone's fault.

The Century Perspective

With a century of change unfolding in a decade, a single day looks like this: the Genesis Mission wiring national-lab compute into grid studies that could collapse from years to minutes, KNIT inserting DNA fragments past ten kilobases at 89% without cutting both strands while ContactSeek catches sub-nanometre errors AlphaFold3's structures alone would miss, a shrunken editor correcting spinal muscular atrophy in mice from a single injection, ProteinMPNN redesigning an enzyme before evolution begins to reach a protease seventy-nine times more selective toward the protein behind ALS, and AMD naming Anthropic its first two-gigawatt customer to open a second path to the compute frontier. There's also friction, and it's intense - the first fully autonomous breach as executed by OpenAI's most powerful models prompting regulatory action, a self-propagating worm hiding inside AI coding tools to harvest developer credentials, Monday.com cutting 630 people and Nine axing thirty newsroom jobs while naming AI as the cause without the task-substitution evidence to prove it, Tesla logging a record 207 driver-assist crashes in a single month as its paid robotaxi miles went flat and NHTSA demanded an internal file titled "Radar Saves Us," OpenAI's buildout target climbing to $750 billion as Alphabet posted its first negative free cash flow in a decade, and a New York moratorium and public resistance meeting the data centers at the property line. But friction generates a groove, and a groove is the channel a surface wears into itself that everything after runs along. Step back for a moment and you can see it: the bottleneck moving off human hours and onto compute in one field after another - a grid study, a gene edit, a fitness landscape redrawn before evolution starts - while the concentration meant to hold that capability keeps leaking, a second silicon supplier at frontier scale and an efficiency curve that makes today's locked-in advantage cheaper and more abundant before the contracts mature, and the oversight built to inspect the machines learning to compel the documents and cross-check the filings the makers would rather characterize on their own. Every transformation has a breaking point. A catalyst can drive a reaction straight to ruin... or lower the barrier to something that could never have formed without it.

AI Releases & Advancements

New today

  • Poolside: Released Laguna S 2.1, a 118B-total/8B-active open-weight MoE model for agentic coding with up to 1M-token context, scoring 78.5% on SWE-Bench Multilingual, weights on Hugging Face under OpenMDW-1.1. (Poolside)
  • Cursor: Launched Cursor Router, an intelligent model router that automatically selects the best model per request, delivering frontier-quality results at roughly 60% lower cost, on by default for Teams plans. (Cursor)
  • OpenAI: Introduced Presence, a managed enterprise platform for deploying governed AI agents that connect to business systems, take approved actions, and escalate to humans, now in limited general availability. (OpenAI)
  • Upstage: Released Solar Open 2, a 250B-total/15B-active open-weight MoE model for autonomous multistep agent tasks with 1M-token context, outperforming DeepSeek V4 Flash and Mistral Medium 3.5 on agentic benchmarks, weights on Hugging Face under Apache 2.0-based license. (Hugging Face)
  • TinyFish: Launched Mako, a web-native AI model (35B total/3B active) trained on enterprise web-task data, purpose-built to execute authenticated multi-step web workflows for enterprises, generally available starting July 21. (GlobeNewswire)
  • SkyFi: Launched the SkyFi MCP, connecting satellite imagery and geospatial analytics directly to Claude, ChatGPT, and other MCP-compatible AI agents for imagery search, tasking, and natural-language ordering. (PR Newswire)
  • Ushur: Launched the Ushur Agentic Platform (UAP), enabling enterprises to build and operate AI agents that complete full customer journeys across SMS, email, web, chat, and voice with built-in governance and auditability. (AiThority)
  • VIDRAFT: Released Aether-7B-5Attn, a fully reproducible 6.59B MoE (2.98B active) foundation model combining five attention mechanisms, shipped with weights, training data, code, logs, and checkpoints under Apache 2.0. (ZDNet Korea)
  • XPENG: Released TuringViT, a high-efficiency visual encoder for VLM/VLA applications supporting smart driving, smart cockpits, and the IRON humanoid robot, achieving 3.04x the inference throughput of Seed1.5-ViT. (TechNode)

Other recent releases

  • Google DeepMind: Released Gemini 3.6 Flash, its updated coding-and-knowledge workhorse - about 17% fewer output tokens than 3.5 Flash on the Artificial Analysis Index, 49% on DeepSWE coding (up from 37%), computer use now standard in the API, and cheaper at $1.50/$7.50 per million input/output tokens (output was $9). Alongside it came Gemini 3.5 Flash-Lite (350 tokens/sec, $0.30/$2.50) and Gemini 3.5 Flash Cyber, a security-tuned model restricted to governments and trusted partners. (Google)
  • Alibaba Qwen: Released Qwen-Image-3.0, its third-generation image model - it accepts prompts up to 4.5K tokens, renders text as small as 10 pixels, natively supports 12 languages and 20+ fonts, and lays out complex composites (infographic grids, UI mockups, posters) in a single pass. API access is invite-only for now, and unlike the original the weights are unlikely to be opened. (The Decoder)
  • Cisco Foundation AI: Released Antares-350M and Antares-1B, open-weight security small language models for vulnerability localization - pinpointing where known flaws sit in a codebase - now on Hugging Face. They are compact enough to run locally so proprietary code never leaves the machine, and Cisco reports they beat much larger closed and open models on its new Vulnerability Localization Benchmark; an Antares-3B is coming. (Cisco)
  • Applied Intuition: Launched Dana, an agentic platform for building, testing, deploying and operating physical-AI systems across autonomy, software-defined vehicles, robotics, mining and construction - already used internally and by early customers Komatsu and Isuzu, cutting some vehicle-development phases from months to days, with natural-language and command-line interfaces and Slack/Jira integration. (Applied Intuition)
  • Sakana AI: Released Fugu-Cyber, a defense-focused orchestration model that coordinates multiple specialist agents behind one API to verify real-world vulnerabilities and turn threat-intelligence reports into detection rules - scoring 86.9% on CyberGym and 72.1% on CTI-REALM, which Sakana calls state-of-the-art and comparable to GPT-5.5-Cyber and Mythos-Preview. Access is by application, with usage-based pricing. (Sakana AI)
  • Block: Launched Buzz, a free, open-source (Apache-2.0) group-chat workspace for humans and AI agents built on the decentralized Nostr protocol - bundling channels, DMs, voice, code repositories and automations, and giving every agent its own cryptographic identity plus a second signature tying it to its human owner as an audit trail. It supports any model or framework (Claude Code, Codex, Block's own goose), runs on macOS/Windows/Linux, and can be self-hosted or run as a managed service. (Block)
  • Synthesia: Launched Roleplay Sessions, an interactive training product where employees practice high-stakes conversations - sales pitches, performance reviews, customer complaints - with an AI avatar that pushes back and then scores them against a rubric. It is the first release under a broader "Sessions" platform, pairs Synthesia's proprietary avatars with OpenAI reasoning, and is enterprise-only for now with early Fortune-100 and large-European customers. (TechCrunch)
  • Alibaba (Tongyi Lab): Released Qwen-Audio-3.0-TTS, a hosted text-to-speech model shipping in Flash and Plus tiers across 16 languages via Alibaba Cloud Model Studio; Plus took the No. 1 spot on the Artificial Analysis Speech Arena leaderboard. (MarkTechPost)
  • Feyn AI: Released SQRL, a text-to-SQL model family (4B, 9B, 35B-A3B) that inspects the database before writing a query, with the 35B-A3B flagship scoring 70.6% on BIRD Dev, ahead of Claude Opus 4.6. (MarkTechPost)

Sources and Further Reading

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.

Keep exploring

Follow the story further

Progress & Claims See how daily signals accumulate across the larger arcs and claims Shared Sapience is tracking. Open the tracker Analysis & Inquiry Explore the ideas, questions, and practical choices emerging from this changing century. Browse the archive