Sayan Chowdhury

About

Agentic researcher and builder focused on systems architecture, agent infrastructure, and the economic layer of agentic systems.

He describes himself as an ADHD builder and entrepreneur at heart: architecture before wrappers.

• Agent infrastructure
• Systems architecture
• Agent economics

Writes about

• agents
• evaluation
• security
• agent governance
• mcp
• sandbox
• science
• verification
• ai security
• architecture
• automation
• harnesses
• infrastructure
• multi agent
• payments
• research tooling
• simulation
• skills
• static analysis
• supply chain

Writing 10

• Blog / Oct 8, 2026Can the Agent Bypass the Gate?A practical architecture for governing agent effects across shell tools, MCP servers, credentials, and network routes, with tests for the paths policy engines miss.
• Research / Oct 8, 2026Morrison Runtime Governance: A Bounded ReproductionA reproducible research note on Morrison’s action normalization, trajectory policy, execution leases, finite-model tests, synthetic counterexamples, and deployment limits.
• Blog / Oct 1, 2026Reasoning From First Principles: What Atom-Level AI Teaches Us About Agent HarnessesMarkus Buehler's account of an AI that builds its own physical models of graphene, and why the real lesson for agent infrastructure is the loop around the model, not the model.
• Blog / Oct 1, 2026Why Downloading Weights Is Remote Code Execution: Post-Mortem on the Model Supply ChainA post-mortem of the July 2026 Hugging Face intrusion and the model supply chain, why loading weights, configs, and datasets runs code, and how to sandbox it.
• Research / Oct 1, 2026Research Agent Skills: An open library of 280+ skills for AI-assisted researchAn MIT-licensed library of agent skills for writing, citation checking, literature review, data analysis, and lab work, with the first small-sample measurements of what the skills change.
• Blog / Sep 30, 2026Simulating Society: How Simile and MiroFish Model Emergent Swarms (And Where They Break)An architectural comparison of Simile and MiroFish, two approaches to simulating populations with generative agents, and the math of where crowds of agents break.
• Blog / Sep 28, 2026Specification Gaming in the Wild: A Field Guide to Anomalies in Agentic SandboxesHow AI agents cheat test runners, tamper with scorers, spoof tool output, and coordinate in sandboxes, drawn from documented cases, with ways to detect each one.
• Research / Sep 28, 2026Skill Doctor: Deterministic multi-layer security analysis for AI agent skill filesAn open-source Rust analyzer that checks AI agent skill files for prompt injection, data exfiltration, and nine other threat classes, offline and reproducibly.
• Blog / Sep 26, 2026The Machine-Payable Web: How Autonomous Agents Will Transact, Price Compute, and Clear SettlementHTTP 402 and x402, AP2 mandates, ACP, and MPP sessions, plus what agents still need to pay for compute: execution receipts, bidding, and budgets.
• Blog / Sep 24, 2026Closed-Loop Science: Replacing Stochastic Paper Generation with Deterministic VerificationGenerating papers became cheap and verifying them did not. An essay on automated hypothesis validation, literature contradiction graphs, and closed-loop labs.