Shawn

take time,
take risk,
take it seriously

[Shawn @shawnli1874]

@nowazhu 闲鱼攒几台迷你主机或者二手服务器，coolify + github actions + 自建 runner，减少不必要的 CI

[Shawn @shawnli1874]

@YinsenW_ 我没找到，让 codex 帮我找找到了

[Shawn @shawnli1874]

@thsottiaux this ain’t no ship

[Shawn @shawnli1874]

@chunxiangai 实测让 opus 通过 @MulticaAI 编排 issue 给其他 agent，通过 issue 描述、状态、评论来交接任务，效果很好很省 token

[Shawn @shawnli1874]

@SheikhSilicon @Polymarket The current figures are estimates, not final data.

[Shawn @shawnli1874]

@10xmylife Criticism seems sophisticated, and making new things often seems awkward.

应该是来自 https://t.co/be7z4ghJhw

[Shawn @shawnli1874]

@jiayuan_jy Grok build 近期会加入到支持的 runtime 里面吗？

[Shawn @shawnli1874]

@tsybka @PolymarketTrade @Polymarket same story didn't happen in 2025

[Andrew Wilkinson @awilkinson]

Over the last two decades, Google captured intent.

You go there, punch in what you want, and it directs you to where you can get it.

Over time, it received more and more "rent" as a middle-man, by charging advertising fees.

Despite this, if you knew how to play ball with Google (pay them for advertising, jump through the SEO hoops), you could build a pretty solid business.

But now ChatGPT has captured intent.

I no longer use Google Search. Like, ever.

(If you'd told me that 3 years ago, I wouldn't believe you)

And this time around, things are very different.

Because now, if your service is digital, ChatGPT doesn't need to direct users to your business at all.

Instead, it can simply solve the problem for the user.

This destroys huge numbers of businesses.

First, it killed education and reference sites.

Need a recipe? It creates one.

Want to learn to code? ChatGPT will teach you inline.

But hell, why even learn? It can now just code for you.

Right now, it's killing code editors and IDEs.

We all thought Cursor was a great business just a year ago.

Now, tools like Codex and Claude Code may render Cursor and Windsurf irrelevant.

Like a stack of punch cards to feed into a mainframe.

You used to Google "how to do my taxes" and get sent to TurboTax.

Soon, ChatGPT will just ingest your accounting data and do your taxes for you.

Next up, will be design and creative tools.

Do I need Photoshop, Figma, or Ableton Live when I can simply type a string of text describing what I want and have it spit out a perfect song, video, image or design?

Earlier this week, ChatGPT welcomed "Apps" into ChatGPT.

Companies like Booking .com, Spotify, and Figma excitedly announced that they were integrating.

I'd wager a bet, that within 3-5 years, OpenAI will simply replace these companies with its own inline services.

After all, the only reason Google didn't do this was because didn't make sense from a resources perspective.

They owned the intent and could have competed—it would have made business sense—but they had limited human capital.

In the old world, competing with a business line required allocation of limited human capital. But now, for companies like Anthropic and OpenAI, it's simply allocation of compute.

Sora is just a small preview.

With one fell swoop, they've created a serious competitor to TikTok.

But it's not just TikTok that is under siege...

Right now, Sora is only producing 10 seconds at a time.

Soon videos will be 1 min, then 10 min, then 60 min, then 2 hours.

In time it will compete with every video platform—TikTok, YouTube, Netflix, and even Hollywood.

Software is the most vulnerable.

They've already started with consumer and soon will move to enterprise.

Adobe. Notion. Salesforce. DocuSign. Zoom.

Any software company that is JUST software or digital assets will be under siege from OpenAI in the long-term.

Building a pure software company today is like opening a Blockbuster in 2007.

You will soon find yourself competing with a juggernaut that has infinite scale and zero marginal cost.

Grammarly? Built into the base model.

Expensify? "Scan and process these receipts."

Calendly? "Find a time that works for everyone."

Notion? "Organize my thoughts on this project."

Stock photo providers like Getty? "Make me an image with X and Y."

Resume builders? "Write my resume for this Google PM role."

Travel planning sites? "Plan my entire Japan trip with bookings."

You get the idea.

So, who is safe?

Businesses that have a hardcore moat.

Amazon is safe because warehouses and logistics are atoms, not bits.

Spotify is safe because it has rights to the music libraries.

Airbnb is safe because they own relationships with millions of property owners.

Stripe is safe because payment processing requires trust and regulatory compliance.

Apple is safe because of its hardware ecosystem and userbase.

The pattern is brutal and clear: if you exist purely in code, you will be replaced by code.

If you have physical assets, regulatory barriers, network effects with real humans, or IP rights – you might be safe from the steamroller for a while.

But unlike the shift from desktop to mobile, which created new opportunities, this shift is consolidating power.

The largest network has always won. But now, they're collecting your profits instead of the rent.

[Shawn @shawnli1874]

@TheCinesthetic Ichiban, nail polish for men

[Shawn @shawnli1874]

@diabrowser pls add spaces and folders

[GREG ISENBERG @gregisenberg]

shipping fast vs shipping smart

There's this weird thing happening in my founder friend group lately where everyone's obsessed with shipping velocity.

Like, genuinely obsessed.

I'll be on a call with someone and they'll be like "Yeah we shipped 12 features this month" and there's this pride in their voice, like they just told me they ran a marathon.

But then I ask "Cool, which ones are people actually using?" and there's this pause.

And I get it. The AI tools make it sooooo easy to ship now. You can go from idea to live feature in like 6 hours with Cursor, Bolt, Lovable, V0 and Claude. The dopamine hit of seeing something you built in production is addictive.

I did the same thing when I first discovered no-code tools in 2019. I was building like 3 apps a week just because I could. Felt like I was crushing it.

But through that I learned there's shipping to learn something and shipping because you feel like you should be shipping.

The difference is huge.

Shipping to learn looks like: "We think our users want XYZ, let's build the simplest version and see if they actually use it."

Shipping because you can looks like: "We haven't shipped anything in a week, let's add that feature we talked about."

When you're optimizing for output instead of outcome thats a trap. You ship five features in a month and wonder why their metrics didn't move.

The ability to build fast became a drug, and now a trap is chasing the wrong high.

The teams that are actually winning have one thing in common

They document everything. Not for users, for themselves. What they're shipping, why they're shipping it, what they expect to learn.

sometimes thats in words, sometimes that's in founder-led video.

This creates this weird insight flywheel. Writing forces you to articulate your assumptions. Those assumptions become testable. The results feed back into better decisions.

The act of writing "we're building this because we think X" makes you realize how often you don't actually know what X is.

The friction to ship got so low that the friction to think feels relatively massive. It's easier to just build something than to sit with uncertainty about whether it's the right thing.

But the best builders I know have gotten really good at sitting with that uncertainty. They'll spend days just talking to users before writing a single line of code.

They ship fast when they need to learn something. They ship slow when they already know what they're building.

So if you're feeling the pressure to ship constantly

Ask yourself: am I shipping to learn or am I shipping to feel productive?

Both are fine. Just know which one you're doing.

[Prajwal Tomar @PrajwalTomar_]

I’ve vibe coded 18+ MVPs for clients using Cursor.

Security was the one lesson I learned the hard way.

Here’s the checklist I wish I had from day one ↓

[Prajwal Tomar @PrajwalTomar_]

This is the most efficient Cursor workflow I’ve used.

If you’re building MVPs fast, this combo is unbeatable:

TaskMaster + Cursor + CodeRabbit

Bookmark this.

[Shawn @shawnli1874]

@dotey 之前有图片 IP 侵权检测的业务，测试了一圈海外服务商和模型之后发现反而是直接提示词+图片输入给 LLM（gemini 2.5 flash & flash pro 差不多）问他检测到了哪些侵权元素，再 json 输出，成本和效果都是最佳的

[Shawn @shawnli1874]

Maybe the difference is whether the system runs on explicit programming rules for everyone, or if it adapts to your unique context and intent. Most traditional software is built on fixed logic that applies to all users. Agents, at least the way people use the term now, are supposed to act based on your specific situation, learning from your data and preferences. Just plugging in an LLM somewhere isn’t enough—what matters is if it can actually personalize its actions for you, not just follow a universal script.

[Shawn @shawnli1874]

@lijigang 也是出于成本，因为同一套界面往往既要服务刚刚注册的新用户（目的是理解），也要服务已经熟练使用的老用户（目的是效率），并且多数时候要根据后者的习惯优化。无需新手引导最好的例子可能是塞尔达，也是因为可以确保每个用户的旅程基本是一样的。有了生成式 AI 之后这个成本限制可能会逐渐被消解。

[Shawn @shawnli1874]

@natiakourdadze done

[Shawn @shawnli1874]

Real change happens when AI tools seize responsibility, remove the human in the loop, and earn enough trust for users to fully let go.

[Hiten Shah @hnshah]

You introduced a new AI tool to your team.

Maybe it helps with writing. Or planning. Or summarizing meetings. You shared it. You explained how it would save time. You got a few nods of approval. And then… nothing.

No one used it. Or they used it once and went back to how things were.

You started wondering if the team just wasn’t ready. Or if the tool was overhyped.

But the real reason is simpler.

You added AI, but you didn’t take anything away.

That’s the root of the issue. You added something smart to a system that was already full. You bolted automation onto old habits. You expected speed without rethinking the system.

When nothing gets removed, nothing feels different.

Teams have operating systems. Workflows. Rituals. Cadence.

And most of those weren’t built to move at machine speed. They were built to coordinate humans.

You bring in an AI meeting note-taker. It writes solid summaries. Pulls out the action items. Organizes follow-ups. But the team still takes their own notes. Still rewrites the recap. Still follows up manually. The AI doesn’t replace anything. So it just adds more.

You give the team access to an AI writer. You tell them to use it for updates and docs. They try it. The draft comes out decent. But they don’t trust it. They rewrite it. Edit every line. Spend just as much time as before, maybe more. So the AI becomes a fancy autocomplete instead of actual leverage.

It’s not because the tool is bad. It’s because the team’s relationship to the work hasn’t changed.

This is the part most people miss.

They believe that adoption is about access. Give people the tool, show them how to use it, and they’ll shift. But people don’t change habits because a tool is available. They change habits when expectations change.

If you tell your team “This AI is here to help,” they’ll treat it like an assistant. Optional. Disposable.

If you say “This AI now owns the update,” the dynamic flips. Now it’s the default. The tool is responsible. The team adjusts to the new normal.

This only works if you take something away.

Cancel the meeting the AI replaces. Stop writing the doc it’s meant to generate.

Set the rule: if the agent can respond, don’t tag a person. That’s when AI starts to matter.

Teams don’t need more tools.
They need fewer steps.

They need clarity around who owns what, especially if that who isn’t human.

Trust is another lever.

If you assume the AI will get it wrong, you’ll end up redoing the work. If you assume it will get it right unless proven otherwise, you’ll actually let go.

That’s the moment real productivity starts. Not when the tool works, but when the team stops trying to hold onto the way things were.

You don’t hire a person, give them no responsibilities, and expect them to create value. The same is true for AI. It’s not useful until it owns something.

So if your team isn’t using that new AI tool, don’t look at the tool. Look at what stayed the same.

The standups are still on the calendar.
The updates are still being written.
The process is still manual.

Adding AI without subtracting anything guarantees it won’t stick. The friction will stay. The rituals will survive. And the tool will become just another tab no one opens.

If you want it to matter, assign it a job.
Remove the thing it’s meant to replace.
Write a new rule. Make it the default.

Then get out of the way.
That’s when it sticks.
That’s when things change.

[Pirate Wires @PirateWires]

NEW: Replit founding data scientist and engineer @giansegato: “I’ve begun to see product managers developing business financial models; designers writing commercial ads; barbershops building custom booking systems; and restaurant owners creating advanced pricing tools... These people always had it in them to do these things. The key difference is that it no longer takes years to learn how.”

Gian’s in Pirate Wires today with a piece on the "unraveling of credentialism" — and how having an edge in the market is no longer about knowing how to do one specific thing very well, but about being biased toward making it happen. Link threaded 👇