the dam brief #13

The Headline

the week's biggest story

Cursor gives code review its own run mode

Cursor launched Auto-review Run Mode on May 29, a new way for Cursor to work longer with fewer approval prompts while keeping execution safer. the changelog frames it as a mode for autonomous review rather than ordinary chat, which puts review, approvals, and runtime limits into the product surface.

this follows a month of agent tooling moving from prompt boxes into controlled workflows: memory scopes, model rules, CodeQL checks, usage cohorts, and repository-level switches. the agent stack is getting less magical and more administrative.

(Source)

The Build

tools, models, and dev releases

• Understand-Anything turns code into a navigable knowledge graph - builders can point Claude Code, Codex, Cursor, Copilot, or Gemini CLI at a repo and explore the codebase as an interactive graph instead of burning tokens on blind file walks (GitHub)

• codegraph pre-indexes a local map for coding agents - the repo gives agents a 100% local knowledge graph for fewer tool calls, fewer repeated reads, and less context-window waste on large codebases (GitHub)

• oh-my-pi is a terminal coding agent with a sharper tool harness - the setup includes hash-anchored edits, LSP, Python, browser tools, and subagents, which makes it useful to inspect if your agent keeps breaking files through sloppy patching (GitHub)

• Microsoft publishes an Agent Governance Toolkit - the repo packages policy enforcement, zero-trust identity, execution sandboxing, and reliability patterns for teams moving agents from demos into controlled environments (GitHub)

• Cursor opens its plugin specification - the repo gives builders a concrete way to inspect the plugin model, official examples, and the surface area Cursor wants third-party tools to build against (GitHub)

• aislop catches AI-generated code smells without another model call - the CLI ships 40+ deterministic rules across seven languages, so teams can add a cheap quality gate before agent-written code reaches review (GitHub)

Founder Fuel

funding, acquisitions, and market moves

• Anthropic reaches a reported $965B valuation - The Guardian says the round pushes Anthropic above OpenAI on private-market value, keeping frontier-lab pricing detached from normal SaaS math (More)

• Groq is reportedly raising $650M after Nvidia's $20B not-acqui-hire - TechCrunch says the inference chip startup is shifting more attention from hardware toward serving model responses faster and cheaper (More)

• SpaceX wins $6.45B in Space Force contracts ahead of IPO - the filing detail turns defense revenue into part of the company's public-market story, not just a launch cadence story (More)

• XCENA raises $135M on the memory bottleneck thesis - the South Korean chip startup is betting that AI infrastructure pressure moves from raw compute toward memory bandwidth and system design (More)

Longevity & Health

aging, biotech, and health research

• chronic kidney disease now affects nearly 800M people - ScienceDaily summarizes a global study that puts the silent condition among the world's leading causes of death and a major contributor to heart disease (More)

• CBD shows Alzheimer's inflammation signals in mice - the study points at the brain's immune response as a treatment target, while keeping the human-use question firmly outside the result (More)

• protein traffic jams may help explain brain aging - Stanford researchers used short-lived killifish to connect stalled ribosomes with faulty proteins, harmful clumps, and memory decline (More)

• human organoids reveal a switch for nerve regrowth - Cambridge researchers found a gene network tied to lost repair ability and tested an existing hormone drug in lab-grown brain-and-spinal-cord systems (More)

The World

regulation, policy, and geopolitics

• the European Commission publishes its first startup and scaleup scoreboard - the ESSS tracks Europe's founder ecosystem through policy, financing, talent, and market conditions instead of relying on one funding headline (More)

• Brussels holds an orientation debate on EU-China relations - the Commission readout keeps trade, security, and strategic dependency on the same table for European companies operating across both markets (More)

• Irish households face a hidden datacentre tax, researchers say - The Guardian reports that power costs tied to data-centre demand are becoming a public bill, not just an infrastructure footnote (More)

Go Deeper

long reads, podcasts, and documentaries

• why everyone hates AI data centers - The Atlantic | staff. a podcast episode on the left-right coalition forming around energy, land, water, and local control. (Listen)

• this literary AI scandal changes everything - Longreads | staff. a magazine's response to suspected AI-generated fiction shows how weak the verification layer still is in creative work. (Read)

• should you automate your life? - The New Yorker | staff. a cleaner question than productivity maximalism: which parts of a life get worse when friction disappears. (Read)

• I turned off my phone for a month and used a landline - Longreads | staff. a useful counter-programming piece for anyone treating always-on access as a default feature of adulthood. (Read)

The Colony

events and community

• today: May 30 - Codex Community Meetup in San Salvador - the colony goes west with a room for people building with Codex and agentic coding tools (RSVP)

• next friday: June 5 - build fridays in Cluj, Hamburg, and Kiel - three workbenches, same rule: bring a laptop and move something real forward (Cluj, Hamburg, Kiel)

• next saturday: June 6 - founder hackathon: build fast & get funded - a Hamburg sprint for founders who want the build loop and funding conversation in the same room (RSVP)