spec_layer

Process is the spec. AI doesn’t transform enterprises. A legible enterprise transforms with AI.

[spec_layer @spec_layer]

"Consensus points one way. Conviction takes you another. That’s where freedom begins."

What’s one time you trusted yourself over the crowd?

[spec_layer @spec_layer]

@GergelyOrosz @steipete The bottleneck always moves to what AI can’t compress. Right now that’s attention. Next it’s the constraint that decides where attention should go.

[spec_layer @spec_layer]

@gdb Exactly the same. Just different compilers.

[spec_layer @spec_layer]

You can give an enterprise all the data in the world.
Without the domain model, they're just told answers.
OntologyOS is the difference between knowing and understanding. One compounds. The other doesn't.

[spec_layer @spec_layer]

@PolymarketMoney Optics …

[spec_layer @spec_layer]

@system_monarch A standard @stripe system design interview question

[spec_layer @spec_layer]

The Solow quote David opens with — “We see the computers everywhere but in the productivity statistics” — has a direct 2025 analogue: We see the AI everywhere but in the business outcomes.

[Zara Zhang @zarazhangrui]

https://t.co/RpZ07FjlZl

[spec_layer @spec_layer]

But build the loop like someone who intends to stay the engineer, not just the person who presses go.

[Addy Osmani @addyosmani]

https://t.co/hIe0UX7z6T

[spec_layer @spec_layer]

This is so …

[Gergely Orosz @GergelyOrosz]

“My company did layoffs a few weeks ago.

The main metric for productivity is AI usage (% of PRs with AI assistance) and PR count per week.

People are burnt out after layoffs, but they don’t want to be next so they are doing what’s requested: use AI, inflate PR count….”

[spec_layer @spec_layer]

Measuring engineering impact is hard. So we count what’s easy to count.

[spec_layer @spec_layer]

Palantir created the FDE role because ontology doesn’t build itself.

OpenAI just launched a $4B company around it. Anthropic is embedding FDEs inside financial services.

When AI Labs copy Palantir’s go-to-market, that’s not competition. That’s validation.

$PLTR

[spec_layer @spec_layer]

@amitisinvesting The gap Karp describes is real. Most enterprises don’t have an ontology because nobody encoded their processes first.

You can’t buy organizational knowledge. You have to build it from the inside.

That’s why Palantir created the FDE role. The ontology needs a human bridge.

[spec_layer @spec_layer]

@atmoio The problems AI creates are mostly organizational, not technical.

Models inherit the chaos of unencoded process. Fix the spec first.

[spec_layer @spec_layer]

@StockMKTNewz The means of production in AI is your operational knowledge.

LLMs borrow it. Ontology owns it.

[spec_layer @spec_layer]

@wintrgrnbuffalo Karp is right. LLMs without ontology are autocomplete at enterprise scale.

Process is the spec. Until it’s encoded, AI just inherits the chaos.

[spec_layer @spec_layer]

AI doesn’t transform enterprises. A legible enterprise transforms with AI.

[spec_layer @spec_layer]

In the AI world, process is the spec. And the killer app is that every company is about to become a regulated company whether they want to or not.

[spec_layer @spec_layer]

Finally

[Vlad Tenev @vladtenev]

Robinhood has officially arrived in Canada. 🇨🇦

We’ve closed our acquisition of WonderFi, marking our entry into Canada through one of the most well-respected crypto platforms in the country.

With 1 million international funded customers, Robinhood’s mission is going global.

[spec_layer @spec_layer]

This weekend I was vibing …
But not code - songs and music..

Oh my. I love it. If someone to ask me if I prefer music created by humans, not sure …

[spec_layer @spec_layer]

Agentic interfaces is the new way of operating creating the next level of integrated economy.

What is the AI native architecture of a company operating in this new age?

[amit @amitisinvesting]

$HOOD

ROBINHOOD OFFICIALLY ENTERS THE AGENTIC ERA.

- Robinhood is launching Agentic Trading and an Agentic Credit Card, letting users connect AI agents to Robinhood through MCP servers so agents can trade or make purchases on their behalf.

- Agentic Trading will use a separate account, so the agent only has access to money deposited there, with push notifications and real-time activity/P&L tracking.

- The Agentic Credit Card lets an AI agent spend through a dedicated virtual Robinhood Gold Card with user-set spending limits, optional manual approvals, expense tracking, and 3% cash back.

Really excited for this. In an agentic world, so many financial decision can be automated or prompted to be done in a way that can help someone legitimately trade or just make better decisions around their spending habits. Robinhood will get access to first-party data and as a result be able to then build more products on top of these that help users better engage with their AI suite.

I know the stock hasn't done the best this year, but that doesn't mean execution and launching new products has to slow down. Fintech isn't in favor right now, eventually it will be, and when it does come back...Robinhood will have spent all that time working on new products and services to accelerate their earnings growth long term.

LFG.

[spec_layer @spec_layer]

Why do software developers use dark themes, because bugs are attracted to the light.