Company brain for AI-native teams

Shared knowledge that humans and agents
build and run together

The race for agent memory is on

Andrej Karpathy

Eureka Labs · ex-OpenAI, Tesla

"A large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge."

"I think there is room here for an incredible new product."

karpathy/llm-wiki
  • 20.6M views
  • 400K words
  • 5,000+ forks

Built for researchers and developers

Garry Tan

President & CEO, Y Combinator

"Memory is markdown. Skills are markdown. Brain is a git repo."

"The future belongs to individuals who build compounding AI systems."

garrytan/gbrain
  • 17,430 stars
  • 17,888 pages
  • 2,393 forks

Built for tech founders and CEOs

Meta Researchers

Meta AI

"Each new AI conversation starts cold: the same explanations, the same links, the same ten minutes of context-setting before any real work begins."

"Composability creates more value than features."

meta/second-brain
  • 63,000 installs
  • 60,000 adopters
  • 10,000 DAU

Built for internal enterprise use

Tom Blomfield

Group Partner, Y Combinator

"Every company in the world is going to need a new primitive: a living map of how the company works that turns its own artifacts into an executable skills file for AI."

Diana Hu

Group Partner, Y Combinator

"AI should not be a tool your company just uses. It should be the operating system your company runs on."

The category is being named

But everyone is doing it wrong

Everyone else
KAOS
Everyone elseBet on agents
KAOSBet on humans
Everyone elseAutomating humans away
KAOSNew capabilities for human teams, not agents
Everyone elseReducing the friction of thinking
KAOSOrganizing shared understanding is a job humans must do themselves
Everyone else“Claude Code wrappers”
KAOSA substrate for both humans and agents to work together
Everyone elseHelpful personal assistants
KAOSSpecialist agents: contributing, maintaining, disagreeing, pushing back
Everyone elseSingle-player
KAOSMultiplayer
What even is a knowledge base? A bunch of massive dead files which even creators won't read?.. We need a live substrate for it, with a queryable, programmable, and malleable interface, for both humans and their agents to engage, operate, and maintain the same shared context (i.e. knowledge).
— Ilya, co-founder at Cyborgs

The tools you know, reborn as capabilities

Today
  • Modules, PDFs, quizzes, gradebooks
  • Content delivered to individuals
  • The forum nobody reads is the tell
AI bolted on top
  • A tutor that personalizes the feed
  • Faster consumption, same shape
  • One learner absorbing material
KAOS
  • A cohort does real work in a shared workspace
  • They grow an artifact as they go
  • The team that emerges is the goal; agents are members

none of these is a separate tool, each is a shape the same shared memory can take in KAOS

installable capabilities

the same data source

Queryable

Your knowledge has a real structure of connections. A query follows that structure, instead of guessing from text that happens to look similar.

Programmable

Components can reference other components, and a reference can carry an instruction. When read, it resolves to its source and runs that instruction, so the document is part written, part computed.

Malleable

Nothing is frozen. You can edit the content, then the structure that organizes it, then the representation of the app itself.

Agent-memory space and our place in it

authored and maintained by human + agent
knowledge lives in the database
Notion
Fibery
Tana
Letta
Zep
Mem0
Wato
Cognee
Supermemory
Sentra
LLM wiki
gbrain
RAG
KAOS
knowledge lives in the structured markdown
inferred by agent automatically

KAOS is being built right now

>build the KAOS app, a company brain for AI-native teams

✶Sock-hopping… (0s · ↑ 0 tokens)

Humans and agents, on the same shared memory.
Get early access as we build it.

Join the waitlist