Knowix

@knowixbuilds

Turning scrap hardware into AI agents to earn a DGX Spark. Making tech & AI brands get seen on X. Building with @bricktopians

wetakeover.xyz

Joined December 2012

Replies Articles Media

Pinned
Knowix
@knowixbuilds
9h
SCRAP TO SPARK update #10 a handmade power-supply cable finally arrived for the Scrap to Spark build. a week after i ordered it. it doesn't fit. luckily, the guy is replacing it with the correct one for free, so the build lives to fight another day. this is the messy part of
Knowix
@knowixbuilds
Aug 6
SCRAP TO SPARK update #9 I thought this was it. Build done, fans mounted, cables routed. I was already thinking about what to benchmark first. Boy, was I wrong. The PSU's PCIe cable comes up 10cm short of the board. Ten centimeters. Fine, I have another modular PSU lying
Knowix
@knowixbuilds
3h
xAI didn't buy Cursor for the code editor. it bought the missing layer between a model and finished work. Grok Bot running on Cursor's backend makes the logic obvious: - agent infrastructure - tool execution - persistent work - users already paying to let AI do real work xAI
Grok Bot
@bot
5h
Introducing Grok Bot, now in early beta. Bots are AI teammates that do real work for you. They sign in to your tools, use them just like you do, and come back with finished work.
00:00
Knowix
@knowixbuilds
Aug 10
i'm starting to understand why people are canceling Claude Max. the daily Fable 5 limit can disappear by midday. - not after running multiple instances. - not after some ridiculous batch. just normal work in one session. that's not a max plan. that's a timer with a premium
Knowix
@knowixbuilds
Aug 10
the next AI platform shift won't happen in a data center. it'll happen on your desk. Meta just opened the weights for Muse Glimmer a 30B model built for local agents: - screenshots - tools - failure recovery under 20GB in 4-bit quantization. 24–32GB target hardware. this
Mark Zuckerberg
@finkd
Aug 10
Today we're also opening the weights for Muse Glimmer, a great 30B parameter dense model that can run locally. Soon we'll also release the weights for Muse Spark 1.2, our latest foundation model. Meta is a strong supporter of open source and I'm proud of these releases. Congrats
Knowix
@knowixbuilds
Aug 10
the cloud vs local debate is missing the most expensive line item: your time. cloud charges per request. local charges in: - hardware - electricity - setup - maintenance - model compromises if you run agents all day and already own the GPU, local can win. if you want to ship
dax
@thdxr
Aug 9
the average OpenCode Go user spent $1.14 per day on deepseek flash v4 this past week the dual DGX setup people are running to do the same costs $10,000 it takes 24 years to break even at 10x the usage it takes 2.4 years