teej

You should call it Braille

ChatGPT has a billion users so surely not all of the public hates it

X has more active users than it ever has in its history

Passkeys are auth garbage. Normal users do not benefit from overly complex auth.

I’ve used AWS for almost 20 years and I can tell you it’s more stable than Azure

Depends how many 3090s you have

Is that the claim the OP is making?

The food pyramid was published by the department of agriculture, it’s always been propaganda.

Easier to skim 1000 flies from a single drum than 100 flies from 100 bowls of soup.

It doesn't sound like your firm does any diligence that would actually prevent you from buying a vendor that has security flaws.

Your coworkers were probably writing subtle bugs before AI too.

pg_ system tables aren’t built for direct consumption. You typically have to massage them quite a bit to measure whatever statistic you need.

10% is a large drag on the cap table.

Evals.

In this case, they could have QA'd the changes, they just didn't care.

You apply with a human in the loop.

There's zero danger writing Terraform. The danger is running `apply`.

glp-1 inhibitors turn out to be better for alcohol and smoking cessation than any other drug we have. Truly a miracle drug.

Clickhouse earned that reputation. However, it was spun out of Yandex in 2021. That kickstarted a new wave of development and it’s gotten much better.

Now that's a name I've not heard in a long time.

Early 2000s Adobe was stacked with web technology. They knew where the world was headed, but didn't quite capture it the right way.

Flash, Shockwave, Dreamweaver, Macromedia Homesite, Fireworks, Coldfusion, Adobe AIR, LiveMotion, Actionscript 3.0, MXML, Flex.

They shipped so much software, it's incredible.

You can specify platform specific wheels in your pyproject.toml

    [[tool.uv.index]]
    name = "pytorch-cu124"
    url = "https://download.pytorch.org/whl/cu124"
    explicit = true
    [[tool.uv.index]]
    name = "pytorch-cpu"
    url = "https://download.pytorch.org/whl/cpu"
    explicit = true
    [tool.uv.sources]
    torch = [
    { index = "pytorch-cu124", marker = "platform_system != 'Darwin'" },
    { index = "pytorch-cpu", marker = "platform_system == 'Darwin'" },
    ]

This is classic innovators dilemma.

Coaches and owners are not rewarded for innovation. Fans strongly discourage taking bets that could fail.

And then there’s preparing for the strategy change. Training, practice, and coaching time is extremely limited. How much do you re-allocate to this new approach? You don’t just tell players to take more 3s, it’s more complicated than that.

So in traditional innovators dilemma fashion, it’s much easier to follow when you see that the new way works. It’s easier to convince everyone (fans, coaches, players, owners) to get on board when you can point to Steph Curry doing it right.

You’ve sold me, I’ll check it out.

I will!

I watched when it was originally aired, but never came back to it. It didn't come together in the end in a satisfying way.

Arcane was really something. Hard to think of anything like it.

Old as in not made in the last 10 years, so not in the list's window (2015-2025)

I enjoy shows where I get completely engrossed in the world and the story. I love shows that I can fall in love with again on a rewatch. And I want to have lingering thoughts about it when it’s over.

True Detective S1 (2014) is perfect television, but is too old for the last-decade list.

The only other definitive Top 3 is Dark (Netflix)

Other candidates:

- Frieren

- Better Call Saul

- Arcane

- Midnight Mass

- Counterpart (underrated)

- Andor

This is a recreation of a fictional computer program from the excellent Apple TV show - Severance.

The work is mysterious, and important.

Season 2 is going now. It’s one of my top 3 shows of the last decade, highly recommend it.

Why this behavior emerges is an active area of research. What they did is use reinforcement learning, this blog post replicates those findings. The “recipe” is detailed in the R1 paper.

My mental model for chain-of-thought is not “reasoning”. It’s more of an iterative search through the latent space of the model.