raza (76/100 fits)

making things, feeling it all, telling my friends i love them.

product @formacity, prev @_buildspace

[raza (76/100 fits) @AlmostEfficient]

how to have love in your heart:
1. forgive yourself
2. forgive your parents
3. forgive the world

if u get stuck, see #1

[raza (76/100 fits) @AlmostEfficient]

@nileshrthr @heyclicky let's goooooo nilesh!!!

[raza (76/100 fits) @AlmostEfficient]

@formacity brimaxxing innit

[FORMA @formacity]

It's time to be Bullish on Britain 🇬🇧

[raza (76/100 fits) @AlmostEfficient]

@KelseyTuoc the freedom of not having to think about limits/costs changes the way i approach work itself

no limit/cost anxiety = more experimentation, risktaking, ambitious moves, no waiting for the 5h limit

hard to go back tbh

[FORMA @formacity]

This is the sign you've been waiting for.

[FORMA @formacity]

Introducing Forma Bristol

Our first ever popup village in the UK

And the permanent home for our new campus

[raza (76/100 fits) @AlmostEfficient]

@farhaj damn that guy in the black shirt has a massive smile i wonder what his secret is

[FORMA @formacity]

Popups are the new startups.

We spent the last 2 years traveling around the world hosting popup villages with @solana

Announcing our next chapter. We're going permanent, and we chose the UK 🇬🇧

Here's why:

[raza (76/100 fits) @AlmostEfficient]

@tszzl im solving this problem by automating things that prevent me from going outside

[raza (76/100 fits) @AlmostEfficient]

im solving this problem by automating things that prevent me from going outside

[roon @tszzl]

automating the computer has made the computer radically more fun and its even harder to go outside now

[raza (76/100 fits) @AlmostEfficient]

fun side effect of LLMs is that you can now parallelize many cognitive styles cheaply.

it would take me months to learn a foreign reasoning style and unlearn my ingrained intuitions. way easier to prompt a model to "think" a certain way

my own pocket roundtable of nutjobs

[QC @QiaochuYuan]

spooky implication that there is potentially some whole universe of "shadow math" that you have to make inhuman mental movements to access so no human have done so yet, that is going to be increasingly revealed by frontier models

[raza (76/100 fits) @AlmostEfficient]

@FarzaTV @heyclicky where did you get these screen recordings of my computer from 10 years ago

[Blair Dulder CPA™ 🧃 @runaway_vol]

vibecoding is fentanyl for people with an iq over 120

[raza (76/100 fits) @AlmostEfficient]

@MitcheIl lmao did an agent write this?

[raza (76/100 fits) @AlmostEfficient]

i forgot i set up custom claude instructions so it knows what time it is and it tells me to go to bed after 8 pm lol

gnite claude. cya tomorrow.

[raza (76/100 fits) @AlmostEfficient]

looks like we're heading towards private, local, fine-tunable models on consumer hardware

your iphone will learn to talk like you at levels of detail you can't imagine

[Paolo Ardoino 🤖 @paoloardoino]

Tether AI breakthrough

Tether AI team just released new version of QVAC Fabric to include the World’s First Cross-Platform BitNet LoRA Framework to Enable Billion-Parameter AI Training and Inference on Consumer GPUs and Smartphones.

Background
Microsoft's BitNet uses one bit architecture to dramatically compress models.
Traditional LLMs operate on full-precision computation, where weights are stored as complex, high-resolution numbers. The innovation of BitNet is that it shrinks these weights into a tiny ternary range of only -1, 0, and 1. significantly reducing memory usage and computation.
LoRA, is a parameter-efficient fine-tuning technique that reduces the number of trainable parameters by up to ninety-nine percent.
Together they slash memory and compute requirements. Yet BitNet has mostly been limited to CPU or CUDA NVIDIA backends, and lacked the support of LoRA fine-tuning.

Enters QVAC Fabric: the unlock
Today, with QVAC Fabric LLM, is the first time BitNet LoRA fine-tuning and inference work cross-platform across GPU vendors and operating systems using Vulkan and Metal backends.
That means support for AMD, Intel, Apple Metal and also Mobile GPUs.
And for the first time ever, BitNet inference runs efficiently on smartphones using mobile GPUs.
On flagship devices, GPU inference is 2 to 11 times faster than CPU while using up to 90% less memory than the full precision models.

The biggest unlock: QVAC Fabric LLM support for BitNet LoRA fine-tuning on heterogeneous GPUs. Our team was able to demonstrate this by fine tuning models up to 3.8 billion parameters on all flagships phones such as Pixel 9, S25 and iPhone 16 and up to 13 billion parameter models on the iPhone 16.

Github repositories:
https://t.co/UxvXHYpRiy : general QVAC Fabric codebase
https://t.co/dFlbXK1kuQ : specific QVAC Fabric's BitNet knowledge base, architecture docs and pre-built binaries

What does it mean?
What used to require dedicated GPUs now runs on consumer hardware.

This breakthrough is the first real-world signal of a local private AI that can truly serve the people.
And this is just the beginning.
In the next months and years Tether will relentlessly continue to invest significant amounts of resources and capital to continue to research and develop open-source intelligence that can scale and evolve on local devices, providing maximum utility and privacy to its users.

The era of Stable Intelligence has just begun.
Free as in freedom.

[raza (76/100 fits) @AlmostEfficient]

this means that LLMs like Claude will soon contain tools like Excel within them. no more tool calling, the retry loops become invisible.

think of it like learning to drive vs needing to call an Uber. just go wherever you want.

[Christos Tzamos @ChristosTzamos]

1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers.

We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy

[raza (76/100 fits) @AlmostEfficient]

claude/codex are pushing my brain in ways it's never been pushed. they have infinitely larger working memory & virtually zero switching cost

i have to load so much context into my head to make a decision. feels like my brain is pushing my eyeballs out

we are the bottleneck

[raza (76/100 fits) @AlmostEfficient]

new things are coming out faster than i can adapt them into my workflows. half of them are outdated in a month

feels like "new tools specialist" should become a real job. need someone whose full-time job is keeping up for a niche + advising you on what'll fit you

[raza (76/100 fits) @AlmostEfficient]

@spottedinprod banger ship lads, i turned it into a claude skill so sonnet can find pattern inspo as we're building
https://t.co/4veGgiXCn3

[raza (76/100 fits) @AlmostEfficient]

i turned spotted in prod into a claude/codex skill

just describe what you want and it searches across 747+ production ui patterns

better results than scrolling loads of clips