XZ Tie

Engineer turned founder. Building @crafto_ai. Exploring AI, content creation, storytelling, and human creativity. Sharing what I learn.

[XZ Tie @XzTie]

OpenAI DevDay 翻车，世界果然是个巨大的草台班子。

今天回看了OpenAI DevDay, 本来想看他们准备的大惊喜，没想到围观了一轮现场翻车。

- Dots demo一直显示"Checking..."
- Codex 线程报错
- Voice 几次504
- https://t.co/8OE8caC4h8怎么跳转到Elon Musk的Grok Bot了，他手真快啊哈哈

[XZ Tie @XzTie]

@pcshipp There is also instinct…

[XZ Tie @XzTie]

@EternitiesAI Yeah - I'm getting a bit tired of all these new agents. They all have very similar features and none of them stood out with a unique moat to me...

[XZ Tie @XzTie]

The personal agent wave really went from a niche experiment to a crowded race in just a few weeks.

- Meta Muse: launched Sept 8, hit #1 on US App Store, with 2.8M downloads in its first 12 days.
- OpenAI Dots: launched Sept 29 (just two days ago), connects to 4,000 apps and is reachable from ChatGPT, Slack, and Teams.
- Instinct: raised two rounds in August, closed its Series C at $10B in September, with 100k+ users and 10% daily growth,
- Grok Bot: beta launched Aug 11 for SuperGrok and Cursor subscribers, reaching 418K weekly users by Sept 14.

And Google Gemini Spark, DoorDash text-to-order agent, , Apple Siri AI, and Manus Cue are all pushing into this space.

I’ve used a few of them. They saved me $50 on groceries and clothes, which is real value.

But I also stopped after the first few days. They make mistakes. Still feel like a ChatGPT wrapper, and handing over access to email, payments, messages, and accounts holds me back a bit.

The question isn’t whether people will try personal agents. Millions already are.

The question is whether they will keep using—and paying for—them once the novelty wears off.

Are you still using these personal agents?

[XZ Tie @XzTie]

Had a chat with my fellow founder friend about the hardest part of entrepreneurship. She is one year ahead of me on this journey.

𝗪𝗵𝘆 𝗱𝗼 𝘀𝗼𝗺𝗲 𝗽𝗲𝗼𝗽𝗹𝗲 𝗾𝘂𝗶𝘁 𝗶𝗻 𝘁𝗵𝗲 𝗺𝗶𝗱𝗱𝗹𝗲, 𝘄𝗵𝗶𝗹𝗲 𝗼𝘁𝗵𝗲𝗿𝘀 𝗸𝗲𝗲𝗽 𝗴𝗼𝗶𝗻𝗴 𝗳𝗼𝗿 𝟱, 𝟭𝟬, 𝗼𝗿 𝗲𝘃𝗲𝗻 𝗺𝗼𝗿𝗲 𝘆𝗲𝗮𝗿𝘀?

A lot of people love the “𝗳𝗼𝘂𝗻𝗱𝗲𝗿” title—it seems to come with freedom, total control, and eventual success. But that illusion fades quickly in a few months, and you realize entrepreneurship is mostly uncertainty, tedious product iterations and bug fixes, user interviews, rejection, and a lot more unsexy stuff.

I liked one analogy from our conversation:

𝗦𝗰𝗵𝗼𝗼𝗹 𝗮𝗻𝗱 𝗰𝗼𝗿𝗽𝗼𝗿𝗮𝘁𝗲 𝗹𝗶𝗳𝗲 𝗮𝗿𝗲 𝗹𝗶𝗸𝗲 𝘀𝗽𝗿𝗶𝗻𝘁𝘀. You have a clear goal—an exam, a score, finishing a task, getting a raise or promotion. You work hard for a few weeks or months and get the result. Good or bad, there’s always a feedback and a reference point. The path is super predictable.

𝗘𝗻𝘁𝗿𝗲𝗽𝗿𝗲𝗻𝗲𝘂𝗿𝘀𝗵𝗶𝗽 𝗶𝘀 𝗹𝗶𝗸𝗲 𝗵𝗶𝗸𝗶𝗻𝗴 𝗮 𝟭𝟬𝟬𝟬-𝗺𝗶𝗹𝗲 𝘁𝗿𝗮𝗶𝗹.  There’s no score. No one is judging you. You can take as long as you want—or quit anytime. You don’t know what you’ll encounter along the way, and no one tells you what you’ll gain at the end. You define what you gain from the journey—or maybe you gain nothing.

But if you want to continue, there's only one thing you can do: 𝗸𝗲𝗲𝗽 𝗽𝘂𝘁𝘁𝗶𝗻𝗴 𝗼𝗻𝗲 𝗳𝗼𝗼𝘁 𝗶𝗻 𝗳𝗿𝗼𝗻𝘁 𝗼𝗳 𝘁𝗵𝗲 𝗼𝘁𝗵𝗲𝗿.
One seemingly boring step at a time. Day after day.

Most of us are trained for predictable sprints through school and work. We assumed that if we work hard enough, success will follow.

But entrepreneurship requires something different: being able to tolerate boredom, repetition, and uncertainty; staying in the game when there's no score telling you how well you're doing; finding what you truly care about without constantly calibrating yourself against the outside world or other people’s judgment.

𝗔𝗻𝗱 𝘁𝗵𝗲𝗻 𝗷𝘂𝘀𝘁 𝗸𝗲𝗲𝗽 𝘄𝗮𝗹𝗸𝗶𝗻𝗴.

#Entrepreneurship #Founders #Startups #BuildInPublic #FounderJourney

[XZ Tie @XzTie]

When I first started posting on X and LinkedIn, writing in English was another barrier. I knew what I wanted to say, but I would constantly wonder: Does this sound natural? Is the grammar right? Is this how a native speaker would say and write?

Naturally I turned to AI tool, ChatGPT, Gemini, Claude, you name it. I'd dump my rough thoughts and documents and ask it to polish them. And I felt great: my writing must sound natural!

It worked. The grammar was better, the sentences were smoother, and everything sounded more professional. But as I wrote and read more, I started to feel that my posts didn't sound like me, or how I would naturally speak. There were words I never heard of and expressions I'd never use. Sometimes reading it felt like reading an instruction manual or a report: perfect but missing emotion, whether that's happiness, sadness, bitterness, anger, or regret.

So I started prompting again: "Make it more natural", "Make it sound like me", "Refer to my previous posts and don't polish too much". Sometimes it worked, but often I found myself editing the AI polished paragraphs again and deleting the "AI-sounding" phrases. A tool that was supposed to save me time trapped me in a prompting loop that took more time. And even after all that, I'd still occasionally have people tell me, "This sentence sounds like AI". And I didn't get more views or likes on those posts.

So I started writing more of the posts myself. Posting on social media doesn't always need to be professional or perfect. Most of the time it's personal—my opinions, emotions, word choices, humor, even grammar mistakes. The point is to express what I think and feel in my way. In fact, those little imperfections are part of what makes each of us sound different—and why people want to hear from me instead of another perfectly written AI post.

I still use AI to help me write. But I no longer want it to make my English “perfect.” I want it to help me say what I mean more clearly, and still sound like me.

[XZ Tie @XzTie]

Yesterday was the Mid-Autumn Festival, one of the oldest celebrations for families to gather together in the Chinese lunisolar calendar. What I find most meaningful about this tradition is its enduring ability to bring people together under the same moon, even when they are separated by oceans and thousands of miles. In a world that moves so quickly, that sense of connection and continuity feels especially worth treasuring.

As an immigrant, pursuing a dream in a new country often means living with distance.

Distance from family.
Distance from old friends.
Distance from the life you once knew.

Over the years, I've missed countless holidays, birthdays, weddings, and family gatherings. Some days, I wished I could simply sit down with my family for dinner. Other days, people I cared about were going through difficult moments, and all I could do was listen from thousands of miles away.

Yet in many ways, living with this distance has also made me stronger and reinforced my decision to stay in this country and continue pursuing my own dreams. Through it all, my family and friends have given me unconditional love, support, and belief. Especially my parents. I’m their only child, yet they have always supported my decision to build a life thousands of miles away from them.

These experiences have shaped who I am today. They’ve taught me resilience, made me appreciate the people I love more deeply, and prepared me for whatever challenges lie ahead.

Happy Mid-Autumn Festival! 🌕🥮

Wishing everyone a joyful and peaceful celebration, wherever you are under the moon.

[XZ Tie @XzTie]

You have to be extremely high-agency, with a strong sense of ownership, to succeed in today's world.

Just came across this article by @apatwa7 on his first year at OpenAI. He has several things surprised him, and one that really stood out to me is "Everybody's job is to be full stack".

This reminds me of the discussion on the most important trait for success, and more people are now favoring agency over intelligence, hard work, or experience. Most of us have the latter: skills can be learned, knowledge can be acquired, we can become quite capable in a new area in a relatively short period of time in today's world that new things are emerging everyday.

But what distinguishes us is agency—the mental mode that you don't need or wait for someone else to tell or approve what you do. You approve yourself.

And I think that's also where a lot of creativity comes from.

[Aalok Patwa @apatwa7]

I recently crossed 1 year working at OpenAI.

I reflected on doing good work, company culture, and the 5 biggest surprises from an action-packed year:
https://t.co/hoVOAgHzyL

[XZ Tie @XzTie]

@ivanhzhao Not a Bach expert, but it felt a bit plain and repetitive to me, missing emotional dynamics, highs and lows, compared to human baroque music. Still impressive though.

[XZ Tie @XzTie]

Loved your learning on "Everybody’s job is to be full stack". Agency has way more power than intelligence, work ethic, or experience. Most of us have the latter, but it is the former that distinguishes us - you don't wait for others to approve or tell you what to do, you approve yourself.

[XZ Tie @XzTie]

@himanshustwts Very true! Agency has way more power than intelligence or hard work, and it is the source of all creativity - you don't let others tell or approve you, you should approve yourself.

[XZ Tie @XzTie]

I had 10+ years of experience in tech, but when I opened X, I still had no idea what to post.

This felt a little ridiculous. I had spent years doing research in computer science, scaling up distributed systems at Facebook and Netflix, building AI products at Notion, managed teams, launching products, and learning things the hard way.

Surely I had something to talk about.

But whenever I open X and stare at the empty post box, I would still think: What should I write about?

Something technical?
A lesson from building products?
A thought about AI?
Something personal?

Having 10+ years of experience didn’t make it easier. It almost made it harder. There were too many things I could talk about, but I had no idea which ones were worth a post.

For a long time, I thought I had an idea problem.

Eventually, I realized: 𝗜 𝗱𝗶𝗱𝗻’𝘁 𝗹𝗮𝗰𝗸 𝗶𝗱𝗲𝗮𝘀. 𝗜 𝗷𝘂𝘀𝘁 𝗱𝗶𝗱𝗻’𝘁 𝗿𝗲𝗰𝗼𝗴𝗻𝗶𝘇𝗲 𝘁𝗵𝗲𝗺.

Interesting ideas rarely arrive labeled: “Here is a great X post.” 😅

They show up in a conversation with a coworker.
A mistake I made.
Something I disagreed with.
A question a customer asked.
An article that changed my perspective.
A lesson I learned years ago and now took for granted.

These were all potential posts. I just wasn’t looking at them that way.

Now, whenever something surprises me, makes me think, or teaches me something, I write it down, then decide whether it is worth sharing later.

Maybe the problem isn’t that you have nothing to post.

You just haven’t learned to notice the ideas you already have.

[XZ Tie @XzTie]

The AI model price + quality war is getting intense. Right around the time OpenAI launching GPT-6 Sol & Luna with 50% price cut, Anthropic dropped Opus 5.5 with even better performance and 40% lower cost.

Probably not great news for LLM companies 😅

But great news for developers building on top of these models—and all of us using AI every day.

Better models. Faster. Cheaper.

Keep the war going. 🍿

[XZ Tie @XzTie]

I used to kill most of my post ideas before I even wrote them.

An idea would pop into my head, and within seconds, another voice would say:
Is this useful enough?
Isn’t this too obvious?
Someone must have said this already.

And that was it. The idea never made it to a draft.

This happened especially when I wanted to share lessons of building reliable systems at big tech, building new products at startups, riding the early AI wave in 2022, and now taking the leap to build a 0-to-1 company.

I would think: “𝗘𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝗶𝗻 𝘁𝗲𝗰𝗵 𝗼𝗿 𝘀𝘁𝗮𝗿𝘁𝘂𝗽𝘀 𝗽𝗿𝗼𝗯𝗮𝗯𝗹𝘆 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗸𝗻𝗼𝘄𝘀 𝘁𝗵𝗶𝘀.” So why post it?

Then one day, I posted my struggle of always wanting to build a personal brand but never having enough time to write or design a post. I talked about overcoming my mental burden, publishing that first post, and slowly making posting a habit.

I got dozens of messages from people saying they had the same struggle. Some told me my story help them take their first step and start sharing too.

That changed how I think about posting.

What feels obvious to me may only feels obvious because I've spent years learning it.
A lesson that took me 10 years to learn may be completely new to someone 2 years into their career.
A mistake I've made five times may save someone else from making it once.

I used to put too much pressure on every post to be original, insightful, and valuable to everyone. But maybe that’s the wrong bar. It doesn’t need to be valuable to everyone.

𝗜𝗳 𝗶𝘁 𝗰𝗼𝗺𝗲𝘀 𝗳𝗿𝗼𝗺 𝗺𝘆 𝗼𝘄𝗻 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲 𝗮𝗻𝗱 𝗵𝗲𝗹𝗽𝘀 𝗼𝗻𝗲 𝗽𝗲𝗿𝘀𝗼𝗻 𝘀𝗲𝗲 𝘀𝗼𝗺𝗲𝘁𝗵𝗶𝗻𝗴 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁𝗹𝘆, 𝗶𝘁’𝘀 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝘄𝗼𝗿𝘁𝗵 𝘀𝗵𝗮𝗿𝗶𝗻𝗴.

So now, when an idea comes to me, I first write it down, then decide later.

𝗚𝗶𝘃𝗲 𝘁𝗵𝗲 𝗶𝗱𝗲𝗮 𝗮 𝗰𝗵𝗮𝗻𝗰𝗲 𝗯𝗲𝗳𝗼𝗿𝗲 𝘆𝗼𝘂 𝗸𝗶𝗹𝗹 𝗶𝘁. 😅

[XZ Tie @XzTie]

GPT-6 Luna and Sol just got 50% cheaper!

Just tried both. Quality and speed beat 5.6—and at half price 💰💰

My guess is OpenAI is seeing bigger wave of AI users and workloads on cheaper models. This will make it harder and harder for expensive models to justify their cost.

Migrating my prod workloads now...😆

[Tibo @thsottiaux]

GPT-6 Sol and Luna are out. Not only are they a very significant improvement across the board, but also in writing and general "you know when you try it" quality.

We are also permanently reducing the API price by 50% making both of them viable for a ton of new usecases and making your usage go further too, even on the subscriptions.

And one more thing. We are loading a banked reset into all accounts of our Plus, Pro and Business users. Let's go!

https://t.co/00DRh1sRrO

[XZ Tie @XzTie]

@thsottiaux Just tried 6 Luna, quality and speed beat 5.6 and half the price 💰💰💰💰💰💰

Migrating my prod workloads now...

[XZ Tie @XzTie]

If you are a founder, don’t talk to your users over zoom; talk to them in person.

I did another blind user interview today. I showed a few people the new Crafto website I've been working on over the past two weeks. It has a live demo where you enter an idea and, within seconds, turn it into a polished social media post with visuals.

I expected people to click the big text box, type a sentence, and hit Enter. Instead, two of them first clicked the "Set up Voice and Brand" box, an advanced feature designed to entice people to sign in and use it. So a sign-in window popped up, they stared at it for a second, closed it, scrolled around the website a bit more, then closed the tab.

They never tried the live demo.

Sitting next to them and watching this was a bit painful, but incredibly useful. The path that felt obvious to me wasn't obvious to them. Something I designed to attract users was actually distracting them from the main experience I wanted them to try.

I used to do most user interview on Zoom. They're efficient and convenient, and I still get feedbacks. But sitting next to someone and watching them open the product fresh—where their eyes go, what they instinctively click, where they hesitate, and when they lose interest—tells me much more about how they actually use it.

It took me 30min to drive there and another 45min to get back because of traffic. Worth it.

The most valuable user feedback isn't always what people tell you. It's what they do when they don't know what you're expecting them to do.

Btw, if you're curious, here's the version I'm testing: https://t.co/X5iuLGgyfQ, I will leave it up for a while and launch the fix separately, so you can see the before and after.

[XZ Tie @XzTie]

@CompleteSkeptic Just heard from my friend that he emailed you guys this past Saturday, exploring Jev for their multi-million user workloads. He got a response in 5 min (do you have GTM person 24/7 LOL?).

By the end of Sunday, he has got the whole setup working.

You guys move insanely fast!

[XZ Tie @XzTie]

Jev’s team moves insanely fast.

Just heard from my friend that he emailed the Jev team this past Saturday, exploring using it for high-scale production workloads. He got a response within 5 minutes—apparently they hired GTM team globally to make sure someone is available 24/7.

By the end of the weekend, they had already helped him get the whole setup working.

In the AI era, speed really matters.