Marcus Buffett

@EarnWhere 14 years, absolutely having fun with the new “how well can you wrangle an alien intelligence to do your bidding” meta

How long have you been developing and are you still having fun?

I've never been more thoroughly owned than when I ordered a shake in the company Doordash order, it exploded over the other items in-transit, and @delk called it a highly asymmetric downside play

Best bed-read chess books: chessreads.com/review/alekhin… chessreads.com/review/capabla… chessreads.com/review/bobby-f… chessreads.com/review/the-lif… chessreads.com/review/the-ris… chessreads.com/review/chess-w… @ilan_ruby @SJPressInc @capakhine #chess #chesspunks #chessreads

I can't get Opus to understand a single basic instruction right now Sonnet 5 confirmed for tomorrow

Think about this a lot

If you're a #chess adult improver (below 2300 online) and have never worked with me, I have a great offer for you which extends till the end of December! DM me to know more about the discount and let's train hard and get that rating moving up! #chesspunks #chesscoach

@thdxr Yeah team variables here, not everything needs to be context. My state store is like this too. It's also nice that you can access these things without having to be in a component.

solidjs people - some contexts are entirely global and will never be used deeper in the tree i've always wondered why bother making these contexts at all - can i just make them normal variables/functions that can be imported

@Mukherjea @natesolon Maybe I should have said vanishingly small ratio instead. I think it’s the minority of adult improvers that are, well, improving

@MarcusBuffett @natesolon "Vanishingly small" is absurd, depending on how you describe 'significant gains'. Unless that means going from 2000 to IM, this claim isn't correct at all.

One of the hardest questions with chess improvement is when to change or abandon a plan that seems to make sense but isn't showing tangible results. On the one hand, you don't want to abandon the plan before it had a realistic chance to work. On the other, you want to be open to evidence. And if something isn't working for long enough, at some point that becomes compelling evidence that there's a problem. Here's where it's really helpful to have access to someone with more experience. If you haven't done something before, it's very hard to know what could or should work. But to someone who's already done it, it may be obvious.

@dhh Dying for M4 support over here, along with a handful of people I know. I’m actually excited enough about Omarchy that I’m going to buy a Beelink, but would be even better if I didn’t have to buy new hardware. I get it’s challenging though.

Omarchy has been on an incredible growth spurt over the last few months. We've distributed over a petabyte worth of ISOs in the last thirty days alone, which is enough for 150,000+ new installs. In total, we're well into the hundreds of thousands of ISO downloads. But what I care more about than the numbers are the anecdotes. I'm constantly hearing from people who switched from Windows or Mac to Omarchy and love this beautiful, modern, and opinionated Linux way. This is something the Linux diehards tend to forget. What seems like old hat to them ("Hyprland?! Psh, I ran i3 in 2006!!!") is completely revolutionary to most people who've never used a tiling window manager. If you've only ever used Windows or Mac, Omarchy is like traveling to a different computer dimension — full of strange, wonderful new colors and designs. There's still work to do before Linux is likely to take over the desktop for "normies". But that was never the mission for Omarchy. We've been targeting people who genuinely love computers, don't mind learning how they work, and are willing to invest time in memorizing hotkeys to be able to become much more proficient and efficient machine operators. That's surely a subset of all computer users, but tons of such people exist! And it's not just developers, although they make up the lion's share. I've also seen Omarchy get embraced by designers and writers. Not just because this is such a fresh, new take on what a computer can look and feel like, but also because, as a community, we've dumped the silly gatekeeping that for so long has hampered the adoption of Linux. It was part of the reason I didn't think Linux on the desktop was something for me personally for so long. So happy to have been proven wrong! And now I'm excited to help others come to the same conclusion. Omarchy is for anyone who loves computers ✌️

@Ra1kshit 👀 @chessbookcom

Super interested to put together a group of people that understand chess and tech well. Comment below if you are one, or i know someone in this intersection.

I'll never financially recover from this

Decades of Clean Code obsession has poisoned AI training When you say "simplify this", they hear "move this stuff around to other places" Guaranteed net-positive lines of code via any AI "simplification"

95% of X ads engagement is fake, because advertising posts react on tap start instead of tap end, so they accidentally activate all the time while trying to scroll your feed. This is both terrible for users (me), and advertisers (also me) @nikitabier

Every day we stray further from god

GraphQL is pretty cool until you start making 3,000 database queries in one request

I love this push you're doing recently to spread the word more about SolidJS, it would be amazing to see it used more widely – I used it myself for years and absolutely loved it. It's a shame React has achieved such a stranglehold. I get that people were tired of moving between frameworks, but deciding that the history of frameworks ends with React has left a lot of DX and performance on the table.

I think going 7+ years now on most React competitors the answers here are pretty reasonable. Well beyond where React was when it rose to adoption. Things are different now. Expectations have scaled with the pervasiveness of React. People will always find an excuse to stick with the incumbent. And there is nothing wrong that. I'm concerned that things have been getting way better after a decade, and that this stranglehold is getting detrimental. It isn't just any given framework. The underlying architectures have been revolutionizing outside of React. Not to mimimize what React has been doing but there are other things out there making significant improvements.

Often, the “best” technology doesn’t win. Why? Because it must be *way* better to justify the risks and switching costs. Is it buggy? Is it easy to learn? Is it well documented? Is the ecosystem sufficient? Is our team willing to switch? Is the support team responsive? Is it secure, fast, and configurable? What are the hidden downsides? Can we trust it will be maintained? Sometimes the devil you know beats the devil you don’t.

This "Horizon Beta" model is sort of GOATed. I gave it a prompt to figure out a non-negative loss function for a gamma distribution head, to prevent gradient interference, and it nailed it. Sonnet, Grok, Qwen3 275B, Kimi K2 had all fallen flat on their face on the same task