Zack Jackson (@ScriptedAlchemy) header
Zack Jackson (@ScriptedAlchemy) profile picture

Infra Architect @ ByteDance. Maintainer of @webpack @rspack_dev - creator of #ModuleFederation #auADHD #synesthesia own opinions.

San Jose, CA
Joined June 2018
Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)Photo by Zack Jackson (@ScriptedAlchemy)
Please don’t use rspack, use rsbuild. Only ever use rspack directly if you are porting off webpack.
10
8
2
85
14,794
That spark tweet went viral enough that while hunting for another one. The seller knew who I was 😆
466
Astra 6.2 has impressed me on algorithm design
1
7
753
You can just design compression algorithms
7
1,085
Zack Jackson retweeted
A UI agent run, with the evidence beside it. Midscene fills a GitHub sign-up form and checks the fields—without submitting. Left: the run. Right: the execution report. @poteto what would you want to inspect before trusting a UI agent? midscenejs.com
1
1
547
GLM 5.3 flash 1.7x increase in effective memory bandwidth of the DGX Spark ~60% smaller weights @ BF16 Lossless CUDA kernel needs more optimization. Ideally want at least 2.5x effective memory bandwidth
Might have found a way to shrink weights by about 50% without quantizing. Will have to benchmark decode speeds.
3
3
779
Might have found a way to shrink weights by about 50% without quantizing. Will have to benchmark decode speeds.
1
4
1,397
Added blind review of every model turn to my fork. When t3 uses orchestration in the future turns, it gets the model card from past performances to help it pick.
2
9
1,493
External codex thread into t3, then LCM allows the internal ones to expand on the original external thread from my iPhone ChatGPT app.
2
4
1,205
I added “group invites” to my t3 fork, so multiple models can share a single chat thread and speak to each other and me.
7
1
46
2,617
I have the spark in a exo cluster and the exo cluster wired into sparkDash. So it joins my other macs. Had to build from source, but it worked
1
8
2,132
Just found a DGX spark 128gb for 3500 listed at a pawn shop 20min ago. Called, fully working. Had no idea what it was. Just listed as “mini server” I’ve never bought something that fast in my life.
37
19
9
2,720
283,850
It arrived! Thanks ipawnishop.com for a historic bargain at $3500
23
9
2
309
36,181
COME ON MY GUY 😂
Doubt. Assuming this story is real… you’re asking us to presume that you found a pawn shop forward thinking enough to POST IMAGES ONLINE OF PRODUCTS, but not intelligent enough to look up the eBay going price for something by model? Come on my guy
3
28
6,759
Bro just got the best deal of his life
Replying to @ScriptedAlchemy
Bro is about to get biggest scam of his life
59
79
12
13,327
1,464,996
I’d pay 100k for this within seconds of seeing it
11
1
89
12,074
I can’t believe just how much more even the cheapest solution for vCPU hours cost. Can get 1000 cores provisioned on a VPS for LESS
3
15
3,168
Overheard in SF “my ram dealer has a hookup for gpus as well” It was me
1
12
1,620
We are mobilising €525 million in investment with the new European Chip Design Platform. The goal? To boost Europe’s semiconductor ecosystem and empower a new generation of European chip design companies. The future of Europe’s chip industry starts now: link.europa.eu/f3v3xF
285
34
163
286
449,103
Bot army found another spark for 4700. NOT GOOD ENOUGH Keep searching.
10
1,381