vacation is canceled. back to work nerds
The only control or “guardrails” that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades! The Trump Administration has stopped AI “people” from doing bad, or potentially bad, “things,“ like Dario (Anthropic!), who is now pretending to be a “perfect little angel” - and we will continue to do so! We already have tremendous CRIMINAL and REGULATORY power over these companies! There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS! We are leading China, and all others, and will continue to do so. Conspiracy Theorists, Treasonists, Traitors, and Leakers, BEWARE! Thank you for your attention to this matter! President DONALD J. TRUMP
( TS: Sep 14 2026, 9:58 AM ET )
1
28
A preview for Pro users: a new personal finance experience in ChatGPT.
Pro users in the U.S. can securely connect financial accounts, see where their money is going, and ask questions based on the information they choose to connect.
Your full financial picture, now in ChatGPT.
Nik Holiday retweeted
My analysis for Llama 3.1
1. 15.6T tokens, Tools & Multilingual
2. Llama arch + new RoPE
3. fp16 & static fp8 quant for 405b
4. Dedicated pad token
5. <|python_tag|><|eom_id|> for tools?
6. Roberta to classify good quality data
7. 6 staged 800B tokens long context expansion
Long analysis:
1. New RoPE extension method
Uses an interesting low and high scaling factor, and scales the inv_freq vector - can be computed in 1 go, so no need for dynamic re computation. Used a 6 stage ramping up approach from 8K tokens to 128K tokens with 800B tokens.
2. Training
38% to 43% MFU using bfloat16. Pipeline parallelism used + FSDP. Model averaging for RM, SFT & DPO stages.
3. Data mixture
50% general knowledge
25% maths & reasoning
17% code data and tasks
8% multilingual data
4. Preprocessing steps
Uses Roberta, DistilRoberta, fasttext to filter out good quality data. Lots of de-duplication and heuristics to remove bad data.
5. Float8 quantization
Quantizes weights to fp8 and input to fp8, then multiplies by scaling factors. fp8 x fp8 then output is bf16. Faster for inference & less VRAM use.
6. Vision & Speech Experiments
The Llama 3.1 team also trained vision & speech adapters - not released though, but very cool!
Working on adding support into @UnslothAI!
Uploaded 4bit bitsandbytes quants for 8b, 70b and 405b ongoing to huggingface.co/unsloth
Somebody just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
73
keltan just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
46
Somebody just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
37
Addicted User just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
32
Nik Holiday retweeted
"We used to look up at the sky and wonder at our place in the stars, Now we just look down and worry about our place in the dirt."
Nik Holiday retweeted
"We used to look up and wonder at our place in the stars. Now we just look down and worry about our place in the dirt."
Interstellar was about e/acc's saving humanity from demoralization & deceleration-induced peril, instead securing humanity's place amongst the stars.
We've always defined ourselves by the ability to overcome the impossible.
And we count these moments.
These moments when we dare to aim higher, to break barriers, to reach for the stars, to make the unknown known.
We count these moments as our proudest achievements.
But we lost all that.
Or perhaps we've just forgotten that we are still pioneers.
And we've barely begun.
And that our greatest accomplishments cannot be behind us, that our destiny lies above us.
Interstellar was deeply e/acc.
Nik Holiday retweeted
We are launching new Zapier AI Actions alongside OpenAI's new GPTs as a launch partner!
samhbarton just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
1
29
I think I first started talking about the "Frontend Ceiling" 4 years ago and its still not well understood —
there are great exceptions, but for the vast majority of bootcampers entering the software industry today as Frontend/JS devs and indirectly React/Serverless devs, you are being quietly graded into a lower tier of software engineer that will NEVER be on a track to VP Eng or CTO.
Justified or not (I can argue both sides) this phenomenon is real and if you do not actively test your comfort zones you may find yourself hitting invisible career ceilings.
Acknowledgements (which I'm sure some pple will omit reading in their eagerness to contradict me):
- you can have a plenty successful career without being VPE/CTO
- many situations where a person from frontend/JS background can get there, but either the business doesnt have that deep of backend needs or the person effectively needs to dualclass because (proving the point) frontend isnt sufficient for a senior eng leadership role in the way that backend is.
237
Nik Holiday retweeted
llama.cpp now supports distributed inference across multiple devices via MPI
This is possible thanks to @EvMill's work. Looking for people to give this a try and attempt to run a 65B LLaMA on cluster of Raspberry Pis 🙃
github.com/ggerganov/llama.c…
Timwj just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
33
Tsatsiki13 just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
1
90
Oel just bought me a coffee on Ko-fi! ko-fi.com/W7W3M6Q5A
72














