WHO WE ARE Percepta’s mission is to transform critical institutions with applied AI. We care that industries that power the world (e.g. healthcare, manufacturing, energy) benefit from frontier technology. To make that happen, we embed with industry-leading customers to drive AI transformation. We bring together: https://challenge.percepta.ai - Forward-deployed expertise in engineering, product, and research - Mosaic, our in-house toolkit for rapidly deploying agentic workflows - Strategic partnerships with Anthropic, McKinsey, AWS, companies within the General Catalyst portfolio, and more Our team is a quickly growing group of Applied AI Engineers, Embedded Product Managers and Researchers motivated by diffusing the promise of AI into improvements we can feel in our day to day lives. Percepta is a direct partnership with General Catalyst, a global transformation and investment company. ABOUT THE ROLE As a Research Scientist - World Modeling at Percepta, you'll build the systems that let us understand an operation at its full complexity. Real operations rarely arrive as clean, structured data: the ground truth about the operation lives scattered across claims, clinical notes, call transcripts, contracts, and the tacit judgment of operators. You'll build models and agents that learn to compress this mess into a tractable, continuously-updated representation — effectively a digital twin of the operation — that forecasters, user models, and optimizers downstream can all reason and plan against. RESPONSIBILITIES - Design and build learned world models that compress messy, multi-modal operational data (notes, transcripts, contracts, telemetry) into tractable, decision-relevant representations. - Model the transition dynamics of real operations — how a workforce, facility, or network evolves state-to-state — so downstream systems can run counterfactuals before a decision touches a real person or asset. - Build and validate digital twins of customer operations, benchmarked against replayable, real-world testbeds with defensible ground truth. - Partner closely with the forecasting and optimization teams to ensure your representations are the right substrate for calibrated predictions and for finding optimal actions. - Bridge research into practice by partnering with engineers to deploy world models into live customer environments and push toward end-to-end production systems. You may be a good fit if you: - PhD degree in Computer Science, Operations Research, Industrial Engineering, or Applied Mathematics or have equivalent research/industry experience. - Have depth in simulation or world modeling - Have experience in novel machine learning techniques for control and optimization, including test-time search and reinforcement learning for sequential decision-making. - Are comfortable implementing and debugging large-scale optimization systems, and designing benchmarks with real, defensible ground truth. - Are motivated by impact in critical industries including healthcare, supply chains, energy, and finance. - Have a proven track record of execution. - Are an excellent communicator with both technical and non-technical stakeholders. - Enjoy extreme ownership. - Are passionate about AI's transformative potential ABOUT OUR TEAM As an organization, Percepta is committed to inclusion and belonging, and our people are our greatest investment! We’re working against an incredibly ambitious mission. It won’t be easy, but it will likely be the most fulfilling work of your career. In addition to competitive compensation, equity, 401(k) matching, comprehensive health and family benefits, and lunch every day, we offer a culture built on autonomy and trust. We strongly believe that the boldest solutions come from teams who dream bigger together, and that means bringing in people from every kind of background, experience, and path. We are an office-first team that values the impact of in-person collaboration, but we focus on outcomes over outputs—believing that we are each responsible for solving the "whole problem" to make our customers win. If you don't check every box on this description but you care deeply about the problems we're tackling, seeking the truth, and wanting to own outcomes alongside a team that has your back, then we want to hear from you! We know self-doubt keeps great people from applying, so please don't count yourself out before we've had the chance to meet you. Tackling the hardest problems takes a range of perspectives in the room, and we believe that's how we build something truly game-changing. Our company values are more than just words on a page. They inform how our team shows up every day: OUR VALUES Dream bigger: We have the unique privilege of taking on the most ambitious problems and we should chase them with optimism, responsibility, and genuine belief that we can make it happen. We have to embrace the hard things when no one else will. Heart in the game: What we're doing matters and we have to give a shit. Internally, that means fixing badness when you find it. Externally, it means honoring the trust our customers place in us with their most important problems. This isn’t a 9-5, nor is it a job we’re ever going to monitor your hours. We promise to put work in front of you that matters and in return, we ask you to promise to care. Win for the customer: Everyone is an engineer and the job of an engineer is to deliver outcomes, not outputs. Everything we do—the products we build, the partnerships we launch, the strategy we set—exists to make our customers successful. Delivery is the strategy. Make the call: Organizations are only as strong as the pace at which they make decisions. Everyone at Percepta should feel empowered to commit and shape the ambiguity in front of them. But "make the call" cuts both ways: make the decision and make the phone call. High-agency decision-making only works with high-bandwidth communication and we commit to never operate in silos. Intensity with kindness: We believe in excellence in execution, candor in feedback, ruthlessness in prioritization, and survivalist urgency. We also believe you don't need to be an asshole to deliver on any of this. The trust built through shared kindness and vulnerability is what makes the intensity sustainable. Percepta is proud to be an equal opportunity employer and we evaluate every qualified applicant on their merits and give equal consideration for employment regardless of race, color, creed, religion, sex, gender identity, sexual orientation, national origin, disability, veteran or uniformed service status, age, or any other characteristic protected under applicable law. Percepta provides reasonable accommodations to qualified applicants and employees in accordance with federal and applicable state law.
The AI Roles Continuum: Blurring the Boundary Between Research and...
The rapid scaling of deep neural networks and large language models has collapsed the once-clear divide between "research" and "engineering" in AI organizations. Drawing on a qualitative synthesis...
Everything is ugly, so go build something that isn't — Raiza Martin, Huxe (ex NotebookLM)
AI Research Talks @ Percepta #2 · Luma
Welcome back to Research Talks at Percepta, where leading AI researchers share their work with the NYC community of researchers, engineers, and operators. For…
Human scientists trounce the best AI agents on complex tasks
State-of-the-industry report finds that, despite the limitations of artificial-intelligence systems, researchers have embraced them.
Pacing the Frontier | Gillian K. Hadfield
The Pacing the Frontier letter calls on the US government to support an international effort to build the technical and governance tools needed to protect our option to pace AI development. I and others have been working on the problem of how to build such infrastructure for ten years, including participating in dialogues on AI safety with Chinese academic colleagues during the past three. Here are my suggestions: 1. Don’t rely on off-the-shelf models like FINRA and the FDA which were built for 20th Century single-domain government expertise. They’re not fit for purpose. 2. Don’t act like no-one’s thought about the AI governance problem before. We’ve spent two years refining a regulatory markets design into working legislative language, for example, and it’s now in AI governance bills in five states and in Congress. 3. Don’t try to write an exhaustive set of rules for AGI first. 4. Pick a domain that can achieve widespread global consensus to start. Mine would be recursive self-improvement: models should not build models. Build the technology that verifies that. 5. Focus relentlessly on building flexible verification infrastructure that is able to enforce whatever rules we can ultimately agree on. 6. Don’t assume we already know how to do this and governments can just write tests into law. Technology needs to be built and by the private sector. 7. Don’t wait for the infrastructure to emerge first. The components and people are there and the ecosystem can scale fast with the right incentives. 8. Incentivize large-scale investment in verification technology by building a governance structure and industry funding that creates a market for private verification organizations. 9. Use licensing and public oversight to ensure verifiers are independent of the frontier labs. 10. Protect sovereignty by enabling each government to license its own verifiers from a global market of verifiers recognized by other countries. 11. Leverage the incentive of global trade for models and model services by requiring verification for market access. 12. Just start. Sources in comments.
CEO-Bench: Can Agents Play the Long Game?
Language model agents are becoming proficient executors at isolated, short-horizon tasks such as software engineering and customer service. Yet real-world challenges require a combination of sophisticated skills that remain largely untested in agents: (1) navigating long horizons amid uncertainty; (2) acquiring information in noisy environments; (3) adapting to a changing world; (4) orchestrating multiple moving parts toward a coherent goal. We introduce CEO-Bench, which evaluates these capabilities together by simulating a representative real-world task: operating a startup for 500 days. An agent manages pricing, marketing, budgeting, and many other aspects of a fictional company through a programmable Python interface, operating in the same environment and facing the same challenges as a human CEO. Success demands analyzing noisy, interconnected business databases, translating signals into sound strategy, and coordinating many decisions with programming. The strongest agents write sophisticated code that simulates customer cohorts to forecast future cash and mines negotiation history to uncover hidden customer preferences. Even so, most state-of-the-art models struggle in this environment. Only Claude Opus 4.8 and GPT-5.5 finish above the $1M starting balance, and neither consistently turns a profit. CEO-Bench takes a first step toward measuring the intelligence required to drive sustained, adaptive progress over time.
The state of the tech industry in 2026 - LDX3 New York
Presentations — Benedict Evans
Every year, I produce a big presentation exploring macro and strategic trends in the tech industry. New in May 2025, ‘AI eats the world’.
Bring a Product Manager Mindset to Your Next Engineering Job
The age of AI demands strong ideas, data driven methods, and experimentation
The Geopolitics of AI Infrastructure - Dylan Patel, SemiAnalysis
FF26 – This is not the tech I signed up for! – An approach to permacomputing - Darío Gutiérrez Mori
ScientistTwo: Pioneering the Human Knowledge Frontier with Autonomous AI
Scientific discovery is defined by the ability to identify the boundaries of existing knowledge and venture into unexplored territory. The ultimate vision for AI in science is problem-driven autonomous research: given a fundamental challenge by a human expert, the AI independently navigates the scientific landscape, uncovers theoretical and empirical bottlenecks, and systematically expands the frontier of knowledge. In this paper, we introduce ScientistTwo, a fully autonomous multi-agent framework designed to realize this vision. Specifically, ScientistTwo takes an initial problem as input, establishes state-of-the-art baselines, formulates novel hypotheses, and coordinates specialized agents to orchestrate an end-to-end discovery cycle without human intervention. Moreover, the framework rigorously conducts experiments using diverse datasets and metrics, refines methodologies through automated ablation studies, and validates research findings via a closed-loop simulated peer-review rebuttal engine. To evaluate ScientistTwo's capabilities against the highest standards of human scientific achievement, we benchmark it across papers accepted at top-tier conferences such as ICLR, ICML, and NeurIPS. As a result, ScientistTwo autonomously generates expert-level, publishable papers and fully verified, executable codebases. Its solutions consistently outperform human state-of-the-art models, and achieve higher average review ratings than human-authored papers under automated AI review agents. These results show that ScientistTwo is not merely an assistive tool but an autonomous scientific pioneer capable of pushing the frontiers of human discovery. Project website: https://scientist-two.github.io/
We're releasing Terminal-Bench-Science: a benchmark for evaluating AI agents on research workflows across scientific domains. 🚀 An ongoing community effort out of Stanford University & Laude… | Steven Dillmann | 13 comments
We're releasing Terminal-Bench-Science: a benchmark for evaluating AI agents on research workflows across scientific domains. 🚀 An ongoing community effort out of Stanford University & Laude Institute, built by the team behind Terminal-Bench together with scientific domain experts at research institutions worldwide. 🧪 ⚛️ 🌍 ➕ ⚙️ Terminal-Bench-Science 0.1 has 70 tasks across the life, physical, earth, mathematical, and engineering sciences. Each one is contributed by a researcher porting their own workflows into challenging agentic tasks, on problems they actually care about. 🧠 Anthropic's Opus 5 dominates most scientific domains with an overall pass rate of ~30%, except the mathematical sciences, where Fable 5 and OpenAI's GPT-5.6 Sol take the lead. SpaceXAI's Grok 4.6 is one of the best models for engineering tasks, Z.ai's GLM 5.3 is the best open model, and Kimi (Moonshot AI)'s Kimi K3 is the only model beyond Opus 5 that lands on both cost and token efficiency Pareto frontiers. 🏆 All leaderboard results are live here: https://lnkd.in/gYh99UfU 💻 You can find all tasks on our Github: https://lnkd.in/gejKmXpj Special thanks to our amazing reviewer team Allen Hart, Ivan Bercovich, Joe Janssen, Jiaming H., Steffen Bollmann, Sergey Aganezov, Ben Blaiszik, Karl Krauth, Aaron Feller, Ben Blaiszik, Yiqi Liu, Han Hou, Robert Joseph G., Robert Zhang; our AI research advisors Ryan Marten, Alex Shaw, Lin Shi, Ben Feuer, Mike Merrill, Gian Segato, Jenia Jitsev, Bodhisattwa Majumder, Peter Clark, Thomas Wolf, Braden Hancock, Andy Konwinski, Alex Dimakis; and especially to our co-leads Sanmi Koyejo and Ludwig Schmidt. | 13 comments on LinkedIn
B. Scot Rousse: "Language, Technology, & Care"
How agents are transforming work
A new OpenAI research paper shows how AI agents are transforming work, enabling longer, more complex tasks and expanding productivity across roles.
Building pro-worker AI | Brookings
Daron Acemoglu, David Autor, and Simon Johnson ask: What is pro-worker AI, and how can we build it?
