My 4-year old daughter and I scooted into the Uber in Rio and within 10 seconds Anna was striking up a conversation about the Flamengo soccer team with the driver. “I am Flemengo!”… “Flamengo won!”. Lucky for her (or I guess she’s starting to learn that Flamengo wins often), Flamengo had won the night before. Surprisingly, she kept the conversation going for 30 seconds before throwing me, the sofredor (the sufferer), under the bus — “My dad is Botafogo!”. Botafogo is famous for not being that good, except for that one time in 2024. This month I attended my second Botafogo game at Nilton Santos, this time with two friends @thiago and @pedro, both life-long Botafogo fans. When we arrived, both @thiago and @pedro said hi to childhood friends whom they had met at this stadium, in this particular area of the stadium, when they were kids. The hot dogs were pretty rough, the choppe was on point, and the 19 year old Argentinian midfielder, Montoro, was slicing and dicing. Football is dramatic. On July 19th, I was passionately rooting against Argentina in the World Cup Final, thinking that Argentinians are dirty players, scum of the earth! etc. And then 4 days later, I was passionately rooting for this Argentinian player! Football is a daily event here in Brazil. You can feel it in the air when Flamengo wins, less so when Botafogo somehow wins.
Now, from Argentina to "AI", there has been a lot of talk this week about open source / open weight AI models, the big US AI companies, how the US must “win” in AI…all this discourse about the Chinese etc. It reminded me of a trip I made to icy Pittsburgh in February 2018. I was one of several Apple engineers on a recruiting trip to CMU ( #1 AI program in the US). I remember my hair had frozen as I walked from my Lyft and into the gymnasium. I sat down as an “Apple engineer” to talk to potential candidates. The doors opened and 100+ students quickly formed a line in front of my table. One by one, each student came up to my desk and placed their resume on the table. I would ask something like, “What are you interested in?” or “What area are you studying?”. Every single one of them confidently and quickly responded: “AI/ML”. Their resumes were all unbelievable, “top math student in India” kinds of stuff. They were polite, eloquent, enthusiastic, and motivated. 90% were Indian or Chinese. I thought of this trip recently when Moonshot AI, founded by Yang Zhilin CMU ’19, released the Kimi 3 model on July 16th which competes with the top “frontier” / US lab models. And then, on July 27 as promised, they “open sourced” the model by posting the weights. The narrative of the past month has been the large US labs controlling and gate keeping their powerful models and posting stuff like this video from Anthropic. And then, with much less noise, Kimi enters the picture without the drama. To me, this video is scarier than the announcement of Kimi 3 and the rise of China etc. I’m a patriot and I want the US to “win” generally, but you got to respect the game.
And with that, I will finish this post with a scattering of random updates and thoughts. I stumbled on Billy Strings recently, initially from
this interview. Pretty late to the game, but I made it. For ripping code these days, I’m currently happy with
Parallel Agents
in Zed, running models via
OpenRouter and a bit of Claude Code too. By the way, the
OpenRouter Rankings page
is worth a scan. Lastly, I spent a week in midcoast Maine with the family in July. Maine in the summer is unmatched. I watched Anna play on a small rocky beach for 2 hours straight. iPads are very captivating for kids, but nothing competes with
a beach of rocks and driftwood in Maine
.
We also saw our good friend Greg at the
Craignair Inn. Greg and Lauren have been building this inn for the past decade and it gets better every year. Persistence and enthusiasm.