Figurε from τhe Fuτurε
Joined April 2024
- Tweets3.7K
- Following858
- Followers764
- Likes10.5K
Λcτual κεικ retweeted
e now supports
- nemotron NVFP4
- GLM 5.3 NVFP4
for your GB300 station
We took time and care to make sure our NVFP4 was done right
throughput beat the hell out of everyone
your loss if you own a station and dont use actual.
Actual Computer turns your NVIDIA DGX Station into a department-scale AI factory.
Read the full case study:
actual.inc/company/blog/case…
Λcτual κεικ retweeted
Optimizations and speedups for more model classes: Gemini, Qwen, Nemotron, Muse, and decision models are now in the shipped actual v45.0 client.
actual.inc
Λcτual κεικ retweeted
The team at @actualinc has had early access to Nemotron 3.5 Super VL and used it to show phenomenal performance of the model on DGX Station using actual. It’s been a privilege to work with them these past few months. Their blog is cross linked in the official launch blog 🚀💚
Actual Computer turns your NVIDIA DGX Station into a department-scale AI factory.
Read the full case study:
actual.inc/company/blog/case…
Λcτual κεικ retweeted
I'm so excited to be releasing this engine.
This represents the first of hopefully a long line of work with Nvidia
We didn't put comparison charts in the main post
but rest assured
WE SMOKED EM
this is the best engine for your dgx station.
its not close.
Actual Computer turns your NVIDIA DGX Station into a department-scale AI factory.
Read the full case study:
actual.inc/company/blog/case…
Unicorn incoming
We've developed a new inference engine, called 'e'.
It's the closest thing to alien technology which we have, and pushes 10,500 tok/s aggregate on a single DGX Station with the new @NVIDIAAI Nemotron 3.5 Super VL model, released a few minutes ago.
Λcτual κεικ retweeted
Actual Computer turns your NVIDIA DGX Station into a department-scale AI factory.
Read the full case study:
actual.inc/company/blog/case…
Λcτual κεικ retweeted
correct, but we are working on solving this problem by allowing any computer to think and then trying to cut the distance between software and the physical limits of the hardware
more aggressively, nobody else is doing it this way
actual.inc
For most of history, intelligence has been built behind walls.
A handful of institutions controlled the capital, the talent, and ultimately, who got to participate in shaping the future.
Bittensor introduces something fundamentally different.
An open economy where anyone can compete to produce intelligence, and where the rewards are designed to follow measurable contributions.
A verifiable meritocracy.
Your reputation doesn’t make you valuable. Your work does.
Miners compete to produce useful AI. Validators evaluate their performance, and the network distributes rewards accordingly.
The better the work, the greater the opportunity to earn.
And this is where the flywheel begins.
Capital attracts talent.
Talent competes to produce better intelligence.
Better intelligence creates more useful products.
Useful products attract real-world demand.
That demand can bring more capital into the ecosystem, strengthening the incentives that attracted the talent in the first place.
A recursive economic loop, where human ambition becomes fuel for technological evolution.
And underneath this entire machine sits an economic foundation inspired by Bitcoin.
21 million TAO. Programmed halvings. No premine. No ICO. No reserved VC or team allocations.
An asset with predictable scarcity, powering an open marketplace for intelligence.
Think about the implications.
Instead of a corporation deciding which minds deserve resources, we are attempting to build a system where measurable contributions compete for them.
Instead of intelligence being developed exclusively within the walls of a few institutions, talent from anywhere in the world can participate.
Every participant pursues their own incentives, yet collectively they can contribute to something much greater than themselves.
We are not merely building intelligent machines.
We are building an economic organism designed to reward intelligence for continuously improving itself.
A machine that turns competition into progress, and progress into the incentives for further competition.
Not evolution by command.
Evolution by verifiable merit.
And perhaps, for the first time, the evolution of machine intelligence can become an open economic process that humanity itself participates in.
Welcome to “everything is mineable” era.
– τ
$TAO #bittensor
Λcτual κεικ retweeted
Here is something more official, and it now comes with
1) AI Agent to inspect and help through your problems and logs
2) Report a problem section that sends logs to a medium admins can immediately share with devs
3) Official guides
4) Chinese language support ;)
5) Endorsement from the subnet devs!
trainathome.ai/zh
trainathome.ai (eng)
This quoted post is unavailable.
Λcτual κεικ retweeted
Its officially a banger
so im going to shill my stuff
if you want to use your inference from anywhere and youre sick of configuring vllm
use actual.inc
Λcτual κεικ retweeted
How to create an open, permissionless, immutable, global marketplace for intelligence:
(1) Figure out how to mine for it.
(2) Figure out how to tokenize it.
(3) Figure out how to scale it.
Bittensor is currently working somewhere between (2) and (3). Implementing Gamma Tokens and creating more on-chain incentives for buy+burns will finalize (2), imo. (3) is finalized when some of the chain-bloat issues are resolved so that the subnet cap can be lifted to 256 and beyond.
Once (1) - (3) are complete, then, and only then, should you decentralize the governance and let the community take over (imo, ofc). I do consider this an essential step for true immutability.
It's worth noting there's no one else even at (1).
Spot on.
it doesn't matter how great your protocol is if the underlying cultural core is rotten. at some point there will be a crisis that will kill the dream
signs that it is rotten:
- well-meaning people who voice an opinion that could harm the image of the protocol are intimidated
- the highest integrity people have lost social standing and the sociopaths who are good at faking it have disproportionate influence
- the social layer has split into different camps who are unable to talk honestly and frankly to each other
- binary thinking is the dominant mode of expression
- the most influential factions routinely treat dubious tactics as acceptable if the stated goal is virtuous
unfortunately, i do not believe this is a problem that can be fixed retroactively
preventative care is the only thing to do. and it requires a strong, present, highly ethical, and highly gifted founder