Software engineer making videos about dev gear, local AI, and weird tech experiments
Joined March 2009
- Tweets 13,580
- Following 1,478
- Followers 22,373
- Likes 12,602
2,240 Photos and videos
the next next model is not far behind though. will they take away OpenAI’s model? 🤔
Jun 12
Don’t use Fable 5 to build personal projects
Don’t use Fable 5 to build business projects
Don’t use Fable 5 for anything really, they’ll end up stealing it from you
6
21
3,461
Alex Ziskind retweeted
Jun 13
ChatGPT export gives you a backup, but not a usable archive.
I built a CLI tool that turns a raw ChatGPT export into a structured Markdown knowledge archive:
- one chat = one .md
- project-like grouping
- pinned chats
- trash split-out
- attachment recovery
github.com/anfedoro/gpt_expo…
2
1
8
1,098
Jun 12
this is very true. Claude code sessions are the new browser tabs.
Jun 11
I think the core problem with AI right now is that it's too slow.
while GPT cooks you start another session. And then another whilst that one cooks. Then another cause why not.
Now you're a slopopotamus.
Fast inference is the unlock. Humans are single threaded.
1
2
16
4,055
Alex Ziskind retweeted
Jun 9
Had Claude Fable 5 log network packets and display them as cars on a highway, different car types = different packet types
255
835
14,769
1,339,211
Jun 9
If you are having a hard time switching claude code to fable, need to reauthorize again and manually select it
2
18
1,974
Alex Ziskind retweeted
Jun 9
It's very cool that Apple shipped a 20B parameter on-device.
You can't put 20B parameters in RAM at any reasonable precision. To make it work they are using pretty exotic architecture by today's standards.
A small model predicts from the query (or prompt) which experts to load from Nand into RAM. The key distinction from a typical MoE is that you do this once per query and then generate all the tokens with the same experts (instead of switching the experts for every token).
77
297
3,197
225,303
Jun 8
first thing in keynotes : sorry about liquid glass being too liquid
2
17
1,529
Jun 2
Used @NVIDIAAI Video Search and Summarization agent to find clips for my latest video. where have you been all my life?
4
2
32
2,275
Jun 2
Better picture. And, of course, a waitlist
22
10
303
61,035
Jun 2
Don’t forget that the RTX Spark machines will likely not come with the expensive ConnectX-7 dual network interface. That will also drop the price
Jun 2
There’s no doubt RTX Spark will be premium and likely expensive but assuming everything will start at $5000 because the 128GB memory 4TB storage DGX Spark does is asinine
That’s the highest end spec, it will obviously start significantly lower
16
3
58
9,364
Jun 2
Is Windows a mistake for the new NVIDIA RTX Spark machines? Patience folks. (new video)
Jun 1
DGX Spark GB10’s came in one memory config (128GB) and form factor (mini pc). RTX Sparks will have the full range: 16GB-128GB, and both laptops and desktops!
9
3
63
10,524
Jun 1
DGX Spark GB10’s came in one memory config (128GB) and form factor (mini pc). RTX Sparks will have the full range: 16GB-128GB, and both laptops and desktops!
Jun 1
Will a 128GB RTX Spark laptop be expensive? probably. But there will also be lower RAM variants, down to 16GB which hopefully will bring the pricing down to around MacBook Air levels. ASUS, Dell, and lots of others already building.
50
15
575
208,178