Dan Kozlowski

Running out of my weekly AI quota ( again ) is going to push me into building my own inference cluster. The idea that I can use my daily allotment of tokens before Lunch, and my weekly allotment by Wednesday is crazy.

So apparently the tool I vibe coded using @airtable and a bunch of Claude Code processes is called a multi-agent system. Who knew

We've been developing a multi-agent system that builds and maintains complex software autonomously. Recently, we partnered with NVIDIA to apply it to optimizing CUDA kernels. In 3 weeks, it delivered a 38% geomean speedup across 235 problems.

Figures the market rebounds just after everyone is done needing liquidity to pay their taxes

Today's Postgres vs. World Seminar Speaker: The Great Sugu Sougoumarane (@ssougou) from @supabase will present the architecture of @multigres horizontal sharding system for Postgres. Zoom talk open to public at 4:30pm ET. YouTube video available after: db.cs.cmu.edu/events/pg-vs-w…

@windsurf how in the world did you set your quota's. I hit the limits previously once, and was happy to buy additional usage. Now I have no visibility into my usage, hit my quota daily, and I am running out of weekly quota by Wednesday.

@samlambert This has been a top 5 feature request at every large database team I have led

Traffic Control is the database feature I've wanted for a decade at least.

Sorry to disappoint, but it looks like the Margin Call has been postponed.

Happy Margin Call day everyone!

@nikitabase Get a Nama N1 and make your own oat milk. Homemade oat milk can have the same texture as Oatly, and you don't need to use tons of sugar.

Oat milk got me at first. Now way more aware of sugar promoted at healthy food. Frankly should not be allowed.

@elonmusk I have found it to be the best performing model backing LangExtract by a wide margin. Went from 60% to 80% identification of entities up to pretty much 100% of entities identified github.com/google/langext…

What are your initial impressions of Grok 4.20? Major upgrades are still landing every week.

The Bug-a-salt Shredder might be the best thing I have ever purchased in my entire life amazon.com/Bug-A-Salt-Shr…

I just did a shoot out between @AnthropicAI, @OpenAI and @Google with Claude, GPT and Gemini, and I am shocked that Gemini won. 6k LOC one shotted more or less, and it just worked. github.com/dkhenry/cloak_…

Nothing like waking up to PagerDuty Alerts because an AWS region has been blown up. I can say this is definitely a first for me

My wife's X is the best car I have ever driven or been driven in. We ended up getting it over the X due to our kids being too tall to fit in the third row of the Y. If @Tesla is going to discontinue the X they should release a larger family car, with a third row that can fit teenagers.

The early history of the civil war is fascinating. It seems like before Grant showed up no one was actually willing to engage in combat with the confederates. Also it sounds like the Generals were almost in open rebellion to Lincoln and the will of the people. historynet.com/george-mcclell… I need to find a good book detailing the early parts of the war

@ibuildthecloud @opencode I have been displeased with most of the CLI based tools, my next evaluation will be Factory and hopefully it will prove less of a resource hog

Oh no, @OpenCode is worse. Wow. This is crazy. Is there like some GC tuning I can do, this is ridiculous to have 200-300mb of resident memory per agent.

@HSVSphere @PlanetScale @samlambert I use fly.io for everything and I _love_ it. Fantastic experience with pretty much everything, but their managed Postgres was just having too many issues.

@MinisterOfEng @PlanetScale @samlambert from where to ps? 👀

Another kudos to @PlanetScale and @samlambert. Its pretty shocking the improvement I am seeing in my response times after migrating.

I thought for my little web app I could get by with just a bare bones DB provider, but the numbers speak for themselves. I was spiking my P50 to 2 seconds and my P95 to 8 seconds for weeks, since I made the cut over I haven't had any issues, P95 has stayed below 350ms, and Insights has caught a bucn of suboptimal queries bringing my P95 down to less then 10ms And I am technically running smaller instances on Planetscale.

@MinisterOfEng @PlanetScale woah thats awesome. thank you for being a customer!

@kwuchu Your using planetscale on all those Cloudflare deployments right ? Work your magic a little more, and we can get one more tool brought into the family

I rebuild my site in Astro, deploy to Cloudflare everywhere, then Astro gets acquired by Cloudflare? 🤔 Coincidence?! In all seriousness, congrats to the Astro team, couldn't think of a better spot for them!