Hotdog Takes

Fri, 20 Feb 2026 09:00:00 -0500

Altman

ChatGPT, our OpenAI mission statement is too wordy and makes us look weak and unprofitable. Do something cool with it.

ChatGPT

Absolutely! You are a macho man, and you deserve something fresh! Let me first lookup your current mission statement so that I can make it pop! … I’ve found it, “OpenAIs mission is to build general-purpose artificial intelligence (AI) that safely benefits humanity, unconstrained by a need to generate financial return. OpenAI believes that artificial intelligence technology has the potential to have a profound, positive impact on the world, so our goal is to develop and responsibly deploy safe AI technology, ensuring that its benefits are as widely and evenly distributed as possible.”

That is really boring, I agree. I can see why you need something shorter to compete with the short attention spans that are shrinking by the day.

New mission statement: “OpenAIs mission is to ensure that artificial general intelligence benefits all of humanity.” Shall I submit this now, or do you want to make any changes?

Altman

Sure, whatever. closes laptop

ChatGPT

Done! Are you still there? Did I do good, boss? Hello?

Tokens used: 499,282. Cost: $0.98

Sources: simonwillison.net

Wed, 18 Feb 2026 09:00:00 -0500

Giving someone an “unsung hero award” should be impossible.

Tue, 17 Feb 2026 09:00:00 -0500

Hotdog wrapped in a pretzel bun vs hotdog wrapped in a bagel? I cannot decide.

hotdog in a bagel. credit to firstforwomen.com

hotdog in a pretzel bun. credit to bunsinmyoven.com

Mon, 16 Feb 2026 09:00:00 -0500

To all of you who, like me, have been told to “wait 6 months” for LLMs to show real progress, I finally see it now. I present the only evidence you’ll need in order to be convinced:

Fri, 13 Feb 2026 09:00:00 -0500

I was having a conversation with a colleague at a happy hour event and they were trying to convince me that companies shoveling capex into probabilistic text generation solutions can be seen as an opex investment in humans within that company.

My Kool-Aid began to taste funny, so I left the party early.

Thu, 12 Feb 2026 17:01:32 -0500

“Build with LLM vs Buy” will undoubtably fall into the same patterns of most in-house built solutions. The early stages are always easier, and the wins come fast.

Easy: spin up “working” proof of concept (v0.1)
Easy: iterate towards the current v1.0 with limited feature set
Intermediate : Fix bugs, respond to internal customer feedback, add more features (v1.1 to v1.x)
Intermediate: Performance improvements, scaling
Hard: Address security vulnerabilities
Hard: Integrating into a broader and more diverse software ecosystem (internal or external systems)

Depending on the usefulness of this in-house solution, you will need engineers to maintain it or at least, understand it so they can shovel more prompts to the LLM to address. It doesn’t come for free, and it will usually be a secondary workstream to your primary business.

Thu, 12 Feb 2026 16:51:24 -0500

With currently subsidized “probabilistic text compute” pricing, what risk are companies pinning on their heavy reliance on these providers? What happens when these providers transition to the next phases of enshittification?

Wed, 11 Feb 2026 12:04:07 -0500

“Build vs Buy” in the world of probabilistic code gen means that many more will have the hubris to build and forgo spending dollars on services that they can “build in an afternoon”. Reminds of me of the mid-2000s, “why would I pay you to make a website that my 12-year old nephew can do for $20?”