I am a computer guy. I own a company. I talk to AI all day. I enjoy free time. I use this account to follow a few folks. I’m on Bluesky🦋 way more. #nopolitics
@Brandon14289412 Then the comments were that the time to get to the “finish line” was a little too shor. Their are time extension tokens and bike power boosts but I guess I need to make them more visible.
I'm only able to put time into this on the weekends. Want to help :-)
@Brandon14289412 Hey, first of all, thanks for the positive feedback. Yeah, I thought about it and figured breaking down sign-up barriers and getting feedback would be best. So far, testers told me the original package on GitHub had some problems, so I recompressed it and...
GLM-5.3 can now be run locally!
The 2-bit model retains ~81% accuracy after we shrunk it from 1.51TB to 239GB (-83% size).
Run on a 256GB Mac or RAM/VRAM setups.
GLM-5.3 is the strongest open model to date.
Guide: https://t.co/NLgb3CMB6A
GGUF: https://t.co/Sqwq3xjgo5
Today is a very historical moment for AI video generation
You can now generate AI video faster than you can watch it
Before it'd take let's say 2-5 minutes to generate 15 seconds of video
@fal made a post-trained Minimax H3 variant called Max which is 50x faster than the original but still maintains quality
It generates 15 seconds of video in 9 seconds!
That means you can now do new things like build a perpetual livestream with it that never ends!
OpenAI CEO Sam Altman: "you no longer need to write prompts yourself"
in less than 35 minutes, he explains how to use ChatGPT better than most people
he originally delivered this lecture to Stanford students
my friend sent me the recording
I watched it last night and realized I was using less than 15% of what ChatGPT can actually do
watch it, then read the guide below to build a system that prompts itself
Files API is now live. 📁
🔹 Free to use
🔹 Upload an image once, then reference it by file_id to save request bandwidth
🔹 Reuse the same image across requests—no need to upload it again
Learn more: https://t.co/lLa1Sbyo44
4/n
I have been fine-tuning LLMs for over 2 years now!
Here are the top 12 techniques I'd learn if I were to fine-tune them:
Bookmark this.
1. LoRA
> Freezes the base weights and trains two low-rank matrices as the update, resulting in ~95-99% fewer params to fine-tune.
2. QLoRA
> LoRA on top of a 4-bit quantized base model.
3. Prefix tuning
> Prepends trainable vectors to keys and values at every layer, weights frozen.
4. Adapter tuning
> Inserts small trainable modules between transformer layers.
5. Instruction tuning
> Supervised tuning on (instruction, response) pairs so the model follows directions instead of just continuing text.
6. P-tuning
> Optimizes continuous prompt embeddings through a small encoder, mainly for NLU tasks where discrete prompts are unstable.
7. BitFit
> Trains only the bias terms, ~0.08% of params, and still rivals full fine-tuning on small-to-medium datasets.
8. RLHF/RLAIF
> RLHF trains a reward model on human preference rankings, then PPO against it. The pipeline behind the first ChatGPT.
> RLAIF swaps the human labeler for an LLM judging. RLHF-level quality at a fraction of the cost.
9. DPO (Direct Preference Optimization)
> Skips the reward model and optimizes preference pairs directly with a classification-style loss. Simpler than PPO.
10. GRPO (Group Relative Policy Optimization)
> Samples a group of responses per prompt and normalizes their rewards within the group. DeepSeek R1 ran on it.
11. RLVR (Reinforcement Learning with Verifiable Rewards)
> Replaces the learned reward model with a checker or compiler returning verifiable scores. The free signal behind R1's math and code.
12. Federated fine-tuning
> Tunes across decentralized clients that share only weight updates, never raw data. For when data can't leave the device.
GRPO needs exactly one scalar reward per response. RLVR (13) produces that for free on math and code by running the answer through a checker or compiler.
But tasks like a RAG answer, a support reply, or a summary have no gold label to match against.
The usual fallback is a hand-written reward function scoring faithfulness, hallucination, and completeness.
It takes days to calibrate, rewards the wrong behavior when the weights are off, and breaks every time you add a tool or edit the system prompt.
RULER, implemented in OpenPipe's ART (open-source), solves this.
During training, it passes the N sampled trajectories to a judge LLM, which ranks them relative to each other against the agent's system prompt and returns the scores.
Relative ranking is more stable than absolute scoring, and GRPO normalizes within the group anyway, so the rankings feed straight into the pipeline like with RLVR.
Here's the GitHub Repo: https://t.co/srt6poinAz
(don't forget to star it ⭐ )
I wrote a full breakdown recently on how exactly this works, with the training loop and code.
Read it below.
Apple has confirmed that the following services are down:
App Store
Apple Music
Apple TV
Podcasts
App Store Connect
TestFlight
Apple Podcasts Connect
Apple Music Subscriptions
@elonmusk I always click your posts, hoping they will be about technology, but I'm always let down because they turn out to be political. None of your yes-men will tell you this... Let It Go… Focus on something nondivisive like tech!