We read the research, so your build starts from what works today
Anthropic and OpenAI publish what their models can do and how they build with them. We read all of it, keep up as it changes every few months, and bring what holds up into every client engagement.
Frontier models
AI now creates new knowledge
In the past months, models from OpenAI and Anthropic have settled problems that mathematicians and physicists worked on for decades. The results are published and were checked by leading researchers.
Mathematics, May 2026
An Erdős conjecture from 1946, disproved
An internal OpenAI model found, in one shot, a construction that puts more pairs of points at distance one than Paul Erdős believed possible, which Thomas Bloom calls a $500 Erdős problem. Nine mathematicians checked it and wrote it up. Tim Gowers: had a human submitted it to the Annals of Mathematics, he “would have recommended acceptance without any hesitation”.
Anthropic mathematician Levent Alpöge published an explicit counterexample to the Jacobian conjecture, open since 1939, in three dimensions and more, and credited it to Claude Fable 5. Terence Tao called checking it “an extremely quick verification”.
OpenAI’s paper proves that infinitely many pairs of consecutive primes lie at most 186 apart, down from the previous record of 246. “The proof is due to GPT 6 Astra.” Mathematicians are still reviewing it, and its Lean formalization rests on stated assumptions.
Anthropic reports that Claude Fable 5.2, working largely autonomously, completed a calculation laid out in 2003 that had been too laborious to complete, and found the exact limit of a technique physicists have used for more than 15 years to rule out candidate theories of quantum gravity. Both results are preprints checked by a physicist, not yet peer reviewed.
Claude Mythos 5.2 devised a way to predict the complete structure of viral capsids, the protein shells of viruses, with open-source protein-folding models. In Anthropic’s tests it accurately modeled 28 of the 34 it attempted.
On GPQA, graduate-level science questions, experts with or pursuing a PhD in the field answer 65% correctly, and skilled non-experts with the web 34%. OpenAI reports 96% for GPT-6 Astra on GPQA Diamond. On Humanity’s Last Exam, a test of expert-level knowledge and reasoning, Anthropic reports 74.2% for Claude Fable 5.2 with tools.
The bottleneck is no longer intelligence. It is deployment.
Disproving a famous conjecture or finishing a calculation physicists had given up on is PhD-level work, and on some problems beyond it. On exams written for PhDs, frontier models now score above the experts in the field.
So the question for a company is no longer whether the model is smart enough. It is how to bring that intelligence into everyday work: into your processes, your tools and your data, safely and measurably. Between what these models can do and what businesses use them for today lies a large overhang. Closing it is our work.
From the labs
What the engineers at Anthropic and OpenAI learned
Recent findings from the labs’ own engineering writing that matter most for a company putting AI to work, each with its figure redrawn and a link to the original.
Trusted: deterministic code
Untrusted: sandboxed
GatewayConfigures the agent, separate container
Credential brokerThe agent never sees secrets
Auto reviewA second agent approves each action
LogEvery config change and every turn
AgentIts own identity
Workspace and networkOwn directories, allowlisted hosts
configuresusesevery action
Hundreds of agents, none trusted
OpenAI has “persistent agents deployed across the entire company, including hundreds of OpenClaw agents”, and “over 500 agents that run proactively in the background doing meaningful work”. The agent and the code it writes are treated as untrusted and sandboxed; trusted, deterministic code owns networking, identity and secrets. A second agent reviews every action (“we enforce auto review on all agent actions”), and “we log everything”.
What it means for you
Agents that run on their own inside a company are safe because of the walls around them, not because the model behaves: a sandbox, a credential broker so the agent never sees secrets, its own identity, a reviewer for every action and a full log. We build those walls first.
Promotes key notesBuilds wikisCorrects notesLoaded at start
Agents that tidy their memory offline
Anthropic splits agent memory into layers: raw transcripts, notes Claude writes during a session, derived representations such as wikis, and a small injected context loaded into every session. An offline pass called dreaming checks notes against the transcripts, corrects them and promotes what matters upward.
What it means for you
An agent that works on the same business over weeks needs memory it can trust, so a careful build keeps the raw log as ground truth and keeps the always-loaded layer small and reviewed.
Five OpenAI engineers ran variations of one broad cleanup prompt across repositories: “we merged more than 1,400 pull requests, deleting over 800k lines of code”, part-time over two weeks. The agent gathered its own evidence and explained each change for human review. On one task, “GPT-6.1 Sol Ultra produced 92% as many findings as GPT-6 Astra Ultra at 14% of the API cost.”
What it means for you
The biggest model at the highest effort is rarely the right default. Start small, review the agent’s work, then scale, and compare a cheaper model or effort setting on a slice of the job before paying for the expensive one everywhere.
OpenAI and deepsense.ai evaluated the same plugin in a standalone function-calling loop and inside Codex. “Pass rates differed by 14 and 21 percentage points between the standalone loop and Codex”, in opposite directions for two models. Across three repeated runs the average barely moved, yet “23 of the 108 (21%) flipped between a pass and a fail”.
What it means for you
An agent that passes a test bench can still fail in your real setup, and a steady average can hide unstable answers. We test inside the system the agent will actually run in, repeat the runs, and have a person review a sample of answers.
Anthropic defends agents at three points: the content they read, the model, and the environment they run in. Model defences are strong but probabilistic, and users approved roughly 93% of permission prompts, so the environment is the hard boundary; Claude Code's OS-level sandbox cut permission prompts by 84%.
What it means for you
A careful deployment does not rely on the model or on click-through approvals alone, but limits what the agent can reach with sandboxes, file boundaries and egress controls.
Mar 4Default effort cut from high to mediumTo avoid very long latency
Mar 26Cleanup of old thinking shipsA bug cleared reasoning every turn, not once
Apr 7Effort default revertedUsers preferred higher intelligence by default
Apr 10Thinking bug fixed (v2.1.101)Root cause took over a week to confirm
Apr 16Length limits added to system promptShipped alongside Opus 4.7
Apr 20Prompt reverted; all three resolvedBroader ablations showed a 3% drop (v2.1.116)
Apr 23Usage limits reset for all subscribers
Three small changes, one quality drop
Reports that Claude Code had got worse traced back to three separate product changes between March 4 and April 16, each hitting a different slice of traffic on a different schedule. One bug passed multiple human and automated code reviews, unit and end-to-end tests and dogfooding, and a later ablation found a 3% drop from one system prompt line.
What it means for you
An AI product can regress without any model change, so prompt and harness changes need their own evals, soak periods and gradual rollouts.
MCP via proxyemitEvent / getEventsexecute(name, input)wake(sessionId)
Decouple the brain from the hands
Managed Agents splits an agent into a session (an append-only event log), a harness (the loop that calls Claude) and sandboxes reached as tools through execute(name, input) → string. Any part can fail and be replaced; credentials stay in a vault behind an MCP proxy, out of reach of the sandbox.
What it means for you
Narrow, stable interfaces let a company swap models, harnesses or infrastructure later and recover from crashes without losing the agent's work.
writesreadsproposesreviewsfeedbackbuildsclicks through
Separate the builder from the judge
Agents asked to grade their own work tend to praise it, so Anthropic split the work across a planner, a generator and a separate evaluator that clicks through the running app. Before each sprint the generator and evaluator agree a contract on what "done" means; Sprint 3 alone had 27 criteria.
What it means for you
Agreeing testable acceptance criteria up front and having an independent check against them is the same discipline as code review and QA, applied to agents.
This field changes every few months. These are the patterns, protocols and mental models we bring into every client engagement, drawn from what the frontier labs publish and from what holds up in production.
Legend
Model or agent
Tool or system
Data
Check
Person
Diagrams are our own simplified drawings of publicly documented patterns. Figures are quoted from the named source.
Agent patterns
The agent loop
An agent decides its own next step. It gathers context, acts, checks the real result and repeats, with stopping conditions and human checkpoints to keep control.
Goalfrom a person
Gather context
Take actioncall a tool
Check resultground truth
Done
Repeat until the task is done, a limit is reached or a person is needed
Retrieval-augmented generation in two lanes. Documents are prepared once. Each question then pulls the most relevant passages and the model answers from them, with citations.
Indexing1Documents2Split into chunks3Embed4Search index
Answering5Question6Retrieve passages7Rerank8Answer with citations
Everything the model sees at once competes for a limited attention budget, and recall gets worse as the window fills. The craft is choosing the smallest set of high-value tokens.
An open standard for connecting AI applications to other systems. The host opens one client per server, and each server offers tools, resources and prompts.
MCP hostthe AI application
MCP serverfiles
MCP serverdatabase
MCP serverbusiness app
MCP host connects to: MCP server, MCP server, MCP server.
The agent prepares, a policy decides how risky the step is, and a person approves anything that matters. Routine steps run on their own; accountability stays with people.
Agent1Prepares the action4Executes and logs
Policy2Rates the risk
Person3Approves or rejects
Quality and operations
Evals
You cannot improve an agent you cannot measure. Each task is run several times and scored by graders: code where possible, a model where nuance is needed, people to calibrate.