Подтвердите e-mail

Для публикаций, комментариев, реакций и сообщений подтвердите адрес.

Профиль

Miguel de Icaza

Профиль Vively

🍉Fun Stack Vibing at @xogot.bsky.social Started Xamarin, Mono, Gnome. Prev: Microsoft/.NET/Xamarin/Mono/VSMac/MAIA http://mastodon.social/@migueldeicaza

Good thread

Travis

I think it’s very hard, in this moment in history, to credulously make the case that the CCP is the biggest state-actor threat as regards AI. They might be, I don’t know. But being domiciled in a country that is wantonly *enjoying* ignoring international law makes it a hard position to take. 1/

160

The linked article below is short and useful. Good read.

Paul Nadeau

Anyway, here's my new favorite framework for understanding the Trump administration which sheds good light on how this is work and provides some actionable advice for governments, so it checks both boxes of what the audience is looking for

173

GodotCon is About to start!

2442

Xmas time

stackotter

SwiftCrossUI's Android support has gotten to the point where my sudoku app written with SwiftCrossUI runs on Android without any UI-specific code changes! (I just had to work around a weird HTTPS issue unrelated to SwiftCrossUI)

051

In Mexico we say “estoy más exprimido que un limón de fonda” and I think this is beautiful

1140

In this household we use Chinese models.

2161

Parenting teens:

2130

First they tried regulatory capture by claiming the models were too dangerous. Now that China beat them at their game, they will try regulatory capture via regulation. Same goal, different pitch. From the same OpenAI guy.

2192

OpenAI head of strategic futures posted: “Open-weight models are inherently decelerationist” This is OpenAI’s version of “Linux is a cancer” Both baseless emotional phrases, but one appealed to the masses, OpenAI’s appeals to Silicon Valley sects and telegraphs the Ezra Klein punditosphere.

35011

The end of the Pelican era

Simon Willison

My notes on Kimi K3, plus some thoughts on what we can still learn from the pelican benchmark even while it becomes further detached from how good the models are at the things that matter (like agentic tool calling across longer conversations) https://simonwillison.net/2026/Jul/16/kimi-k3/

080

The "Let me make me a slide deck with Office" workflow: Takes you to "m365" dot "cloud" dot "microsoft" and then shows you a chatgpt wannabe. They lost the plot.

3373

Yesterday I tried to make a PowerPpint and forgot the web site. So I figured “it’s probably Microsoft office”. And Microsoft has really made it hard to reach their office products now. And once you login for the fourth time, you can finally open a document. A cheap AI is running PM over there.

3121

We went from the models being the moat to the apps and services that OpenAI and Anthropic built around them being the moat. They are shallow moats - with startups and chubby companies aiming for them. Time to deploy that cash on PR, lobbying and data centers.

180

Arena.ai reported this. And it is open source, weights to be released on July 27th.

Sung Kim

Kimi K3 (Open Weights by July 27, 2026.) 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost

1545
Показать ещё