Lancer

Building AI infra for deal teams. M&A and finance leader @animocabrands || prev. @morganstanley. @dukeu alum

[Lancer @LiLancer52145]

Predictions:

1) Finance will have a massive agent + GitHub moment. It will move from artifact native (excel, ppt, docs) to state native (what happened, what changed, why it mattered)

2) Tooling will move to agent-native UI. Workbench + agent flow. Think Ramp, Harvey, Cursor

[Lancer @LiLancer52145]

We got GPT 6 before GTA 6

[Lancer @LiLancer52145]

Sometimes I feel burned out from building. Then I read this out loud to myself

[Paul Graham @paulg]

An email I sent to the summer 2007 YC batch to encourage them before Demo Day. It's all still true today.

[Lancer @LiLancer52145]

The epic battle is on: Stripe vs. Ramp

Both want to manage enterprise P&L and monetary + machine flow

[Ramp @tryramp]

We bought https://t.co/4WFUJAf1Cc.

Now we're building the best one.

Our customers buy quadrillions of tokens every month through Ramp, and we've run our own AI on this router for 3 years to keep costs down.

Today it's yours at https://t.co/4WFUJAf1Cc

[Lancer @LiLancer52145]

If you always use AI to write for you, you default to allowing general knowledge and templates to dictate your own voice.

You’d never go above the sea level. And you cannot win at sea level.

[Scott Stevenson @scottastevenson]

Demand for non-slop AI writing would be massive. It’s wild how much we’ve spend training models that can only write longform docs in one terribly cringe style.

[Lancer @LiLancer52145]

Over the past 24 hours, I’ve read a lot of analysis on the OpenRouter acquisition.

5 key points worth diving deeper:

1/ Flow of economic + machine energy:

OpenRouter is effective a clearing house / brokerage for global AI compute. It’s one of the most important flow businesses of the AI era.

Stripe manages the flow of money. OpenRouter manages the flow of machine intelligence.

2/ Winning enterprise P&L:

Stripe already sits on the revenue side of enterprises. OpenRouter sits on the inference-cost side.

Putting these two sides together, Stripe effectively gets a good view across revenue, AI spend, and gross margin, for potentially millions of customers.

It is a gateway for them to manage the holistic P&L. (Watch out Ramp)

3/ Levers around payment economics:

Stripe already processes payments for OpenRouter, so there is an obvious opportunity to optimize some payment economics.

But I think the more interesting possibility is using the scale of Stripe’s developer network to negotiate with model providers. This is to potentially shift some of the 5.5% fee currently paid by users toward the model providers themselves.

4/ The stablecoin angle:

OpenRouter uses prepaid balances and an internal ledger to efficiently handle huge numbers of model calls worth fractions of a cent.

Credit cards are a terrible rail for settling those transactions one by one. Stablecoins are much better suited for this kinds of funding and settlement.

OpenRouter can handle the internal metering; Stripe, Tempo, and stablecoins can potentially handle money moving in and out.

Stripe executives have repeatedly talked about agent-to-agent transactions as a major future trend, so this pathway feels quite natural to me.

5/ Does model routing have moat?

Building a router itself is not terribly technically difficult. I can build a basic version over a weekend by vibe coding.

What is difficult is maintaining reliability, pricing, latency, settlement, and supplier relationships across hundreds of models, every single day, at scale, for tens of millions of users.

That is why I still think this would become a highly concentrated flow business where scale compounds and the strongest players get stronger.

[TBPN @tbpn]

BREAKING: Stripe agreed to acquire OpenRouter for more than $7B, per Bloomberg

[Lancer @LiLancer52145]

Read this and tried Grok Bot and Grok 4.6. I finally came to understand why SpaceX acquired Cursor. My bias was IDE felt last generation and has been replaced by CC / Codex. That judgment is clearly oversimplified

- Cursor has data. Musk has compute (more than anyone else)

- Composer 2.5/3.0 is best on the value-to-performance curve. Grok 4.6/4.7 is already SOTA

- Taste matters. Product shape is the moat: Grok Bot is a good example, so intuitive, lightweight, and elegant

- Good products can catch on quickly and go vertical

- Iterate fast and relentlessly

- Don't bet against the rocket man

[Sarah Wang @sarahdingwang]

https://t.co/hSWR1zQ0iQ

[Lancer @LiLancer52145]

Genius marketing move:

1) Makes Anthropic look terrible. Makes Kimi look responsible

2) People are talking about selling out, which creates more attention

3) People want what they cannot get, which creates more demand

Just imagine the reaction when they reopen the plan

[Kimi.ai @Kimi_Moonshot]

Kimi K3 has received far more love than we expected, and our GPUs are feeling it.

Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected.

We're adding capacity as fast as we can and will reopen new subscription spots in batches.

Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable.

Thank you for your patience and understanding!

[Lancer @LiLancer52145]

Foundational models will be commoditized and price war will continue (including subsidies), until model intelligence becomes such a norm that we no longer discuss it.

Value will accrue to those who know how to best wield such powerful tools. Application layer will rule.

[Scott Stevenson @scottastevenson]

The discovery of LLMs might be like the discovery of fire or electricity: completely indefensible, a commodity spread rapidly to hordes of people who build applications around the foundation.

The vast majority of the value accrual may happen at the application layer longterm.

[Lancer @LiLancer52145]

I still think about this post every once a while. Especially when I read online that people say OpenAI/Anthropic are the last companies to ever exist

It gives me hope

[Scott Stevenson @scottastevenson]

https://t.co/ULlBkT3qye

[Lancer @LiLancer52145]

A prompt that keeps surfacing unexpectedly great ideas while building:

“Forget what I’m already doing. Don’t give me the obvious answer.

If you walked into my company today as Paul Graham, what is the highest-leverage thing I’m completely missing?

What am I not seeing, not considering, or not daring to do?”

[Lancer @LiLancer52145]

@zarazhangrui Notice the full access 👀

[Lancer @LiLancer52145]

Sometimes even if Codex cannot finish a long task within the usage limit, it would work until a reasonable stopping point; then it pauses and tells you what needs to be picked up later.

Genuinely good product design that respects its users.

[Nick Prince🛡 @Nickprince]

codex is goated for finishing what it started after hitting usage limits

[Lancer @LiLancer52145]

Codex talking to each other > Claude bot talking to each other on *Slack*

Just my opinion

[0xDesigner @0xDesigner]

talk to a friend's codex

[Lancer @LiLancer52145]

@0xDesigner is this real? Got me excited for a min

[Lancer @LiLancer52145]

Similarly to this, I’m building a series of agent skills for deal teams. If you happen to work in finance and frequently use agents, I’d love to connect

[Lancer @LiLancer52145]

I review 20+ deals every year, and the first hour is often just cleaning up the deal folder.

So I built a simple skill that turns inbound materials and data room dumps into an agent-ready repo structure for investment and M&A work.

Try it here: https://t.co/SCsmbkYU2Z

[Lancer @LiLancer52145]

The world puts too much weight on skill and not enough on intent.

Met many talented people who are mentally checked out. Talent is surprisingly easy to waste.

With AI, it’s hard to beat someone who genuinely cares, takes ownership, and keeps pushing themselves.

[Lancer @LiLancer52145]

@ben_ai_eng If AI is a multi-decade wave just like the internet, I am hopeful that the most impactful companies haven’t even been founded yet. Long way to go!