JD Mankovsky

Apple geek. Amazing wife, handsome 17 yo twin boys and beautiful daughter. Tesla cars are a technology game changer. Own S , X (2025), Y (2026)

[Greg Joswiak @gregjoz]

Apple Creator Studio just got some exciting updates! Here are a couple great ones:

📽️Final Cut Pro's Generate Caption uses on-device AI to automatically transcribe audio.
🎛️Logic Pro's rebuilt Chord ID picks up chords more accurately for a smoother music production process.

[JD Mankovsky @appljd]

@munster_gene Don't forget the strong residual value on a 3 year old Mac (30% or more). The residual value actually increase recently on older devices.

[JD Mankovsky @appljd]

@elonmusk @t_blom Will it be able to back into our garage ?

[JD Mankovsky @appljd]

@LewisHamilton Thank you Lewis ! You are amazing and we love you !!

[JD Mankovsky @appljd]

@dedene @markgurman To me it’s shocking that Europe has
lost most of its innovators because of all the regulations that make zero sense. All they are looking for is protectionism through regulations and getting access to user data under the same excuse.

[JD Mankovsky @appljd]

@RoKhanna @BernieSanders It always starts with let's taxe the billionaires and then you take everyone ! How about finding ways to use AI to first optimize the very large budgets that are already in place. We don't trust you anymore.

[Shashank Prasanna @shshnkp]

MLX team is 🔥. Apply now!

[JD Mankovsky @appljd]

@giolongoo @ZeClint Private APIs are often used first to make sure they are mature. Once an API goes public Apple does not want to make changes to prevent app incompatibility when you go from iOS N to N+1 (for example). That has been the case for many years. TouchID API was private at first.

[Anemll @anemll]

https://t.co/EYmpueOCk8

Instead of forcing the entire model into DRAM, the full model is stored in flash memory (NAND). Because NAND-to-DRAM bandwidth is too slow to swap weights token by token, as standard MoE models require, AFM 3 Core Advanced makes routing decisions per prompt.
.....
The model periodically reselects and updates the activated experts during the token generation process.

[JD Mankovsky @appljd]

@0xSero Agreed. You need to also mix using local models.

[Greg Joswiak @gregjoz]

Congrats to this year’s Apple Design Award winners! We love seeing the creativity, thoughtfulness, and exceptional experiences developers bring to apps and games each year and we’re excited to see what they dream up next. Can’t wait to celebrate at #WWDC26!

[Greg Joswiak @gregjoz]

All systems glow for a great #WWDC26 next week!

Tune in June 8 at 10 am PT.

[Tim Cook @tim_cook]

Five years. 1,800+ students. And Motor City is just getting started! 

We're proud to celebrate Detroit's Apple Developer Academy and its programs, empowering creatives to strengthen the city's tech ecosystem and build something the world hasn't seen before.

[JD Mankovsky @appljd]

@nathanclark_ @MKBHD I thought I was getting old ! 😉

[JD Mankovsky @appljd]

@Prince_Canuma Finally !!

[JD Mankovsky @appljd]

@milonspace @Merocle Once the local models are on the SSDs you don’t need high speed. WiFi 7 will be just fine.

[JD Mankovsky @appljd]

@AndeanViber @Ex0byt Same question…

[JD Mankovsky @appljd]

@piercefreeman @antirez @ggerganov 😂

[JD Mankovsky @appljd]

@SAKrisT @ClementDelangue @huggingface Apple Silicon chip architecture. Unified Memory. Low power high performance.

[JD Mankovsky @appljd]

@MosheRecanati @NFTCPS I would install LM Studio. Would guide you on the models that would run on your device. They also have an article on how to integrate / have Claude code use the local llm. You need to modify the claude.json file.