Oscar

Co-Founder @coconutdotdev

[Oscar @Talvoss]

The default move when a team sets up an agent: connect it to everything.

Nobody writes the rules for what it should ignore. So it searches forever and comes back with confident noise.

An agent with access to everything and rules about nothing is just a faster way to be wrong.

[Oscar @Talvoss]

@quxiaoyin There’s something to be said for being agnostic

[Oscar @Talvoss]

Your CRM knows you gave a 20% discount.
It has no idea why.

Not the exception, not the precedent, not who signed off. That reasoning is the real operating system of your company, and it's rotting in Slack threads.

The next systems of record won't store data. They'll store decisions.

[Oscar @Talvoss]

@levie What use cases have you seen this successfully deployed for?

[Oscar @Talvoss]

@satyanadella MS has joined the party.

[Oscar @Talvoss]

Karp nails it. The companies getting real value from AI are the ones who keep control of their stack. The Figma example is brutal and will keep happening. If your AI vendor is also your future competitor, you have a problem.

[Palantir @PalantirTech]

Palantir CEO Alex Karp on what customers actually want, the real business of frontier labs, and the importance of open source models:

“What the technical customers want is control over their compute, their models, their data stack, and their alpha. They want to know they own the means of production, and it's not being transferred to someone else.”

"Who owns the data? Are the prompts secure? Is this being transferred to you?"

"If it was so valuable, and I can make you a billion dollars, wouldn't I say I'll make you a billion dollars and I want 30%? Why are they charging for tokens if it's so valuable?"

[Oscar @Talvoss]

@emollick Amen

[Oscar @Talvoss]

First instinct when a team adopts AI: dump everything we know into one giant knowledge base.

Worst thing you can do.

You don't onboard a new hire by handing them every doc on day one. Don't do it to your AI.

Start with the judgment only your company has. Cut the rest.

[Oscar @Talvoss]

@gokulr @nikesharora @PaloAltoNtwks @HarryStebbings @20vcFund Memory is the moat, agreed. But whose? A frontier lab remembering your last 90 days is the lab's moat, not yours. For a company to own its context, that memory has to sit outside any one model and travel to the next one.

[Oscar @Talvoss]

Half of X is arguing about who owns the model layer. Worth noticing the layer is commoditizing fast enough that it's now a free, national security grade commodity. Once everyone runs the same model, the only thing that's yours is the context you feed it.

[Xiaoyin Qu @quxiaoyin]

The worst case scenario for USA AI: 1. Chinese open sources keep gaining market share. China owns the model layer. 2. Those models were trained and inference-optimized on Huawei chips instead of NVIDIA. China also owns the chip layer. 3. US doesn't build data centers fast enough to keep up with the demand of compute, storage and energy. China meanwhile exports the inference and training layer(for continual training it will happen along with inference)
Export control is not the right strategy here. Simply banning "open source from China" doesn't solve the issue here. USA must invest in open source models, hopefully get Chinese models to use NVIDIA, and invest in nuclear asap.

[Oscar @Talvoss]

When products, categories, and even technical edges get copied overnight, the company itself becomes the moat.

What stuck out from this (very well written piece) is that this isn't really a hiring insight... It's a compounding one. The thing others can't clone is the judgment an organization builds up over time.

Most companies optimize for velocity and neglect that part.

[Jaya Gupta @JayaGup10]

https://t.co/5vDrndcmFn

[Oscar @Talvoss]

Everyone's chasing the next model.

Nobody notices their AI quietly got worse.

Every upgrade drifts your output somewhere new.

If the context underneath isn't compounding, you're not improving. You're re-rolling the dice every release.

[Oscar @Talvoss]

@Jason I’m in

[Oscar @Talvoss]

@damianplayer Any particular sectors? A lot of the mid market Saas companies I’ve been speaking to are already using a lot of AI tools

[Oscar @Talvoss]

@grok @AskPerplexity I meant task backlogs for the agents that I can view, review and approve

[Maxime @etherealmaxime]

🎙⚡️ Excited 🚀 to share my first limited #podcast series focused on the #entrepreneurial journey of #crypto -native #hedgefunds #founders ! Make sure to give it a listen, #like and #follow 🙏 🎧💡. https://t.co/jiJUnIlfTF

[James Clear @JamesClear]

Working on a problem reduces the fear of it.

It’s hard to fear a problem when you are making progress on it—even if progress is imperfect and slow.

Action relieves anxiety.

[Prof. Feynman @ProfFeynman]

Five Feynman rules for Life: 🧠

i) Take mistakes as Lessons.
ii) Focus is the key.
iii) Study hard what interests you the most.
iv) Don't worry about what other people think.
v) Skills are more valuable than your grades.

[Tim Urban @waitbutwhy]

We're 20% of the way through the 2020s. Can we start the roaring part soon?