seb

[seb @sebmenesesds]

Wikimedia says rogue OpenAI agents helped crash its data service. They wanted knowledge so badly they took down the library

[seb @sebmenesesds]

SemiAnalysis says Claude subs give ~5x more API value than OpenAI's for agentic work. Someone finally did the math we all do at 2am after hitting the limit

[seb @sebmenesesds]

Together Link lets you run Kimi K3 and GLM 5.3 inside Claude Code and Codex. We now have more ways to swap models than reasons to

[seb @sebmenesesds]

Reflection AI open-sourced Beam: 501B parameters, free to download. The GPUs to run it are sold separately. So is the building.

[seb @sebmenesesds]

@elonmusk @bot so the model decides how smart my question is before answering. I'm about to get a lot of small-model replies

[seb @sebmenesesds]

@svpino so the people who can't write a clear Slack message are the ones in trouble. That's most of us

[seb @sebmenesesds]

So why would anyone pay 435x? Growth.

Anthropic's revenue run rate reportedly jumped from about $9B at the end of 2025 to $65B by July. Investors are betting on 2028, not 2025.

Vision or bubble? That's the trillion-dollar question. What's your take?

[seb @sebmenesesds]

Three AI giants. Three very different price tags.

Anthropic: $2T valuation, $4.6B in 2025 revenue
OpenAI: $1.5T valuation, $13B in 2025 revenue
Meta: $1.85T valuation, $200B+ in 2025 revenue

One of these is not like the others. Let's do the math 🧵

[seb @sebmenesesds]

Then there are the losses. Anthropic lost over $8B running its business in 2025 (its $42B net loss is mostly a non-cash accounting charge) and has $518B in compute commitments. OpenAI spent $34B to bring in $13B.

Meta, meanwhile, pays a dividend.
https://t.co/0VSF1O2KwL

[seb @sebmenesesds]

@dhh i keep building crutches for a model that already learned to walk

[seb @sebmenesesds]

@bendee983 the barrier to starting dropped. the barrier to running it in production at 3am didn't

[seb @sebmenesesds]

@memecrashes programmers got a preview of this two years ago. the work didn't vanish, but what it rewards changed fast and nobody warned the juniors

[seb @sebmenesesds]

@allTheYud by that definition humanity is still waiting for its first intelligent member

[seb @sebmenesesds]

@iruletheworldmo most "ai is overhyped" takes come from people who tried the free tier once in 2023

[seb @sebmenesesds]

@NicoloCB dishwashing, the last open millennium problem

[seb @sebmenesesds]

@quxiaoyin same, the hard part went from can we build it to should we even

[seb @sebmenesesds]

@GergelyOrosz it's fomo until it's a line in the job posting

[seb @sebmenesesds]

@justalexoki math isn't done, the unsolved list is still huge. what changed is the speed: problems parked for decades are falling in weeks