Lam

What was an impromptu decision to come to Argentina turned out to be a pleasant surprise! I didn't expect that my team would be able to deliver the full solution, nonetheless winning one of the finalists prize. After 3 days and 2 nights of little sleep and hours of coding/building/discussing, my team eventually was invited to the stage to pitch in front of 750+ hackers and audience. Quite a surreal experience that I was able to crown @ETHGlobal finalist a 2nd time, even though this hackathon had the most number of ETHGlobal finalist, meaning it was very competitive. I think this moment, taken by a professional photographer, captured the spirit of hacking and what it meant to take ownership. The photo was taken at around ~6:30AM in the morning, just a couple of hours before the final submission was due. At that point, I was still struggling to give the right pitch, and the audio from my mic didn't function properly, but after the 6th recording, I came to an acceptable demo and submitted our project! Thanks to my team (@vaughnhew @ballew @netun0 @Ghost_xD2) that pulled all nighters with me to build this project. Also thanks to all the organizers and sponsors for making this hackathon experience smooth and fun! Until next time in this little Web3 world! Hasta la vista!

@sh1sh1nk Flying to either Japan or South Korea from SEA then onwards to Canada should be ok, I believe.

finding a flight from South East Asia to Canada i see hiked up prices, then, avoiding the flights through the conflict zones and very limited options this gives me dejavu of lockdown days

Someone audited Garry Tan's website after he bragged about 37K lines of code shipped per day. 78,400 lines of AI slop. 6.42 MB just to load the homepage. 169 requests. for a newsletter blog. Shipping velocity is not the same as building something good. What do you actually optimize for when you're vibe coding?

@konotarogomame Chúc ngon miệng! ベトナムに行ったことがありますか?

ベトナム大使と昼食を食べながら意見交換。

One insight from my recent conversation with Bruce Yang (CEO of @agnesai_sapiens): The future of software may not be building tools for humans. It may be building tools for AI agents. If that happens, entire categories of SaaS could change. Really fascinating discussion on the global AI race, accessibility, and what the next generation of AI companies might look like.

Asked my @openclaw assistant about the main highlight at @ns today and this is what she told me

@html5cat @trq212 Smart man

@trq212 good thing my sleep schedule is 5am–11am PT

To manage growing demand for Claude we're adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged. During weekdays between 5am–11am PT / 1pm–7pm GMT, you'll move through your 5-hour session limits faster than before.

@HumansNoContext 1 + 1 = 3 and do it 100x, 😆

Just published my first Substack article about earning crypto on @solana, specifically through @SuperteamEarn and @SuperteamMY. Being surrounded by various folks building on Solana @ns, I dove deeper into this. The process alone has taught me more about the Solana ecosystem than months of passive reading. If you're in Southeast Asia, especially Malaysia, and curious about Web3 but don't know where to start, this might be for you (link in reply).

Rate limited. Frustrated. Opened Airtable to cool off. Noticed it has free Claude built in. Tested it with a tough differential equation and got full solution in 10 seconds. The free tier is hiding in places nobody's looking. What tools are you sleeping on?

Source article: davidoks.blog/p/why-the-atm-…

The ATM didn’t kill bank tellers. The iPhone did. That line stuck with me because it’s basically a product lesson: Big shifts don’t always come from automating the core task. They come from changing the interface around the task so the whole workflow gets redesigned. Feels very relevant to AI right now. A lot of startups are building “AI that does X job.” The bigger opportunity might be rebuilding the workflow around X so the old role structure stops making sense. Where do you think that happens first?

Full read: adlrocha.substack.com/p/adlrocha-int…

The "who has the best model" debate is already obsolete. Models converge. Everyone gets access. What doesn't converge: your proprietary data, domain know-how, and the messy real-world integrations that took years to build. That's the new SaaS layer forming above the LLMs — context + connectors + guardrails. If you're building in a legacy-heavy industry (insurance, banking, healthcare), you're sitting on more moat than you realise. The question is whether you'll move first. What's the context layer in your industry that no one's wired up yet?

Network School (@ns) runs on: Learn. Earn. Burn. Fun. Today was Fun day. Laser tag. Found a corner, stayed quiet, let everyone come to me. Final score: 24,360, nearly 2x 2nd place. Turns out strategic positioning beats running around in circles. Same applies in business — or does it?

Living in a pro-tech community (@ns) and being surrounded by people building startups using AI can make you think that you're late to the AI race. Looking at the picture below reminds me that we're still very early. Most people on earth haven't even used AI yet. Of the people using AI, most are free chat users.

Software ETFs are down 30% since January. Everyone's calling it the SaaSpocalypse. But here's the thing: AI doesn't kill software companies. It kills the ones that were never defensible to begin with. The moats that actually matter (proprietary data, deep workflow integration, customer trust built over years) don't get erased by a better LLM. They get more valuable. The companies dying right now were one good engineer away from being replaced anyway. That's not AI disruption. That's just selection pressure. What do you think survives the next 3 years?

Spent the last few weeks testing free AI courses while building out insurance research pipeline. Most "top AI courses" lists recycle the same 5 names. Here are the ones I actually found useful, by what you're trying to do: Getting started: - Google (grow.google/ai) - Anthropic (anthropic.skilljar.com) Building things: - fast.ai (course.fast.ai): still the best if you learn by shipping, not by lecturing - DeepLearning.AI (deeplearning.ai/courses): great for applications of AI in industries The one nobody talks about: - BlueDot Impact (bluedot.org/courses): AI safety and governance. If you're building AI into regulated industries (insurance, finance, healthcare), this is the gap most builders don't even know they have. What's missing from this list? Add yours below.

@AIDailyGuy Nice product you got there. Thanks for sharing!

Understanding the problem is everything. We understood that B2B teams waste 60% of sales time on unqualified leads. So we built Dashform as the first form builder for agent use — AI creates adaptive quizzes, qualifies leads automatically, routes only the good ones. Close rate up 40%. getaiform.com/blog/dashform-…

Cursor raised $2.3B and crossed $1B ARR. They now produce more code than any other AI agent on the planet. If building software becomes nearly free, the winners won't be the best coders. They'll be the people who understand a problem deeply enough to know what to build in the first place. That's a pretty big shift if you're building products in 2026. What do you think changes first?