dessatel

[dessatel @dessatel]

Here is DeepSeek-R1 running on M4 Pro Mac mini on Apple Neural Engine, llama-3.1-8B model. LUT4, Context 512.
It's amazing it only takes 3.2W on ANE.
Latency is 15 t/s. It seems to emulate reasoning 🤣
It's only 4GB, so it probably will run on an iPhone.

[dessatel @dessatel]

That’s funny 🤣

[Lukáš Ronald Lukács @lucxsronald]

Things that can prevent Pogačar from winning the Vuelta:

1. Crash
2. Illness
3. Heatstroke
4. An alien invasion
5. Roglič finding a time machine
6. Socks being 1 mm too high
7. GPS sending him up the wrong mountain
8. Mihai Simion picking him for every stage and jinxing him
9. UAE mechanics accidentally putting gravel tyres on Pogačar’s bike
10. Pogačar realizing he forgot to press “Start” on Strava, turning back to the start again and missing the time cut
11. Pogačar deciding he loves being a domestique more and helping Almeida instead
12. Organizers giving him a supermarket bike from Carrefour before the final stage
13. Pogačar being disqualified for being too good

📷: alan.bucar #LaVuelta26

[dessatel @dessatel]

Cool
SpaceX trail tonight

[dessatel @dessatel]

SpaceX launch 🚀 tonight

[dessatel @dessatel]

@nandoprince93 What is app used on iPad?

[dessatel @dessatel]

Good listen on AI6 chip, Tesla, xAI
https://t.co/FQrxibDAot

[dessatel @dessatel]

Ouch, Dojo is no more. 🥺 Google TPU won?
https://t.co/xqiweVKxIV

[dessatel @dessatel]

Good CXL overview and update

[𝐷𝑟. 𝐼𝑎𝑛 𝐶𝑢𝑡𝑟𝑒𝑠𝑠 @IanCutress]

I've been waiting for this for years. Finally it looks like anyone can do it, without the need for special tuning or collaboration. A nod to @Marvell and @AMD for making it happen!

https://t.co/CSULmadLSS

[dessatel @dessatel]

@IanCutress @marvell @AMD I believe this would be useful with some MoE variations inference if you can read memory simultaneously from multiple CXL cards

[dessatel @dessatel]

When it’s hot in the Sound, head to the mountains

[dessatel @dessatel]

@ChipsandCheese9 What is total memory bandwidth on this ?

[dessatel @dessatel]

New kid on the block. Pricing is competitive with Cursor.

[Andy Jassy @ajassy]

Introducing Kiro, an all-new agentic IDE that has a chance to transform how developers build software.

Let me highlight three key innovations that make Kiro special:

1 - Kiro introduces spec-driven development, helping developers express their intent clearly through natural language specifications and architecture diagrams for complex features. This comprehensive context helps Kiro’s AI agents deliver better results with fewer iterations.

2 - Kiro features intelligent agent hooks that automatically handle critical but time-consuming tasks like generating documentation, writing tests, and optimizing performance. These hooks work in the background, triggered by events like saving files or making commits. It’s like having an experienced developer constantly reviewing your work and handling the maintenance tasks that often get delayed.

3 - Kiro provides a purpose-built interface that adapts to how developers work. Whether you prefer chat interactions or working with specifications, Kiro supports your workflow while keeping you in control of the development process.

Kiro is really good at "vibe coding" but goes well beyond that. While other AI coding assistants might help you prototype quickly, Kiro helps you take those prototypes all the way to production by following a mature, structured development process out of the box. This means developers can spend less time on boilerplate code and more time where it matters most – innovating and building solutions that customers will love.

Starting today, Kiro is available for free during preview and supports most popular programming languages.

Here’s how to get started with @kirodotdev today: https://t.co/Ne5m2Nh4wC

Excited to see how developers use Kiro, and to work with the developer community to continue to shape Kiro moving forward.

[dessatel @dessatel]

Anthropic is sure ahead in quality and speed of the models, especially if used via Claude Code

[Chamath Palihapitiya @chamath]

Wow.

[Chips & Wafers @ChipsandWafers]

AI Server Growth Continues🚀

June revenues are in for the AI server ODM's, totaling 18.5$ USDb and up +113% YoY.

$NVDA $AMD $INTC $AVGO $MRVL $AMZN $MSFT $META $GOOG $CRWV $NBIS

[dessatel @dessatel]

@corsix @tenstorrent Nice!

[dessatel @dessatel]

Now Intel joins AMD in 1.6TB/s MRDIMM announcements

🤞for M4 or M5U even if it’s not replaceable memory

https://t.co/PLKgkticEh

[dessatel @dessatel]

@fleetwood___ @arcinstitute @grok What about the theory that the cell cannot be fully explained by deterministic science, and that quantum mechanics is truly required? Could this limit the applications of LLMs?

[dessatel @dessatel]

Interesting, @dylan522p on Meta AI, MS/OpenAI, Apple, NVidia, Grok and Super Intelligence 🤔
https://t.co/CenOOwBUUE

[dessatel @dessatel]

@handleym99 @lafaiel For sure. MacBook, low price. It’s just that dropping x86 allows to use iPhone/iPad chips and lower overall cost. It’s probably going to have smaller RAM and battery.

[dessatel @dessatel]

@markgurman Good to see Apple overcoming “not built here”, should buy them enough time to build their own Siri later 👍

[dessatel @dessatel]

@lafaiel That and Apple might be paying some sort of license fee? I bet it was planned years https://t.co/nVaeeJcOq0 should also allow 🍎 to compete against chromebooks