Prompting is my thing.
Earth
Joined January 2023
- Tweets3.3K
- Following1.1K
- Followers3.2K
- Likes6K
MEGA retweeted
I still notice confusion here:
Ask ☀️Sol to start fresh threads in 🌕Luna. Yes, Sol cannot do Luna subagents by default. So literally ask it to start a thread in Luna, Max effort. Also instruct it to have that output be submitted back to the parent Sol, review it, and if it's not good enough, send it back to the Luna Max thread again. Repeat until output is approved.
☀️Sol -> 🌕Luna
🌕Luna -> ☀️Sol
(☀️Sol -> 🌕Luna, 🌕Luna -> ☀️Sol)
repeat until mission accomplished.
codex tip: I always use define-goal instead of /goal. It's basically like having a master level genius AI write your goal better than you could. And it's an official OpenAI skill.
You can get your prompts to run longer and your output better. Ask a high effort Sol to research + plan it. Sol then spawns a new thread for a workhorse model (Luna Max) to execute the work and return a result to Sol for review. LLM-as-judge to determine if Luna's output is acceptable progress toward goal
github.com/openai/skills/blo…
MEGA retweeted
Introducing Netherite!
Minecraft 1.11.2, rewritten from scratch in C and CUDA, bit-verified against the real game.
This is one trained agent playing in it - then the renderer it actually trains through - then 7,200 live worlds stepping in lockstep on one GPU.
MEGA retweeted
How AI is quietly creating a new digital economy.
@IrenaCronin and I write this newsletter every week.
AI is creating far more than smarter models. It's giving rise to an entirely new digital economy built on cloud services, infrastructure, AI agents, subscriptions, and compute marketplaces, where the biggest opportunities may come from the businesses that power AI rather than the models themselves.
Read and subscribe for free: unaligned.io
MEGA retweeted
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: huggingface.co/moonshotai/Ki…
Tech report: github.com/MoonshotAI/Kimi-K…
Tech blog: kimi.com/blog/kimi-k3
MEGA retweeted
We removed ~80% of the Claude Code system prompt for our newest models, this is what we've learned about writing system prompts, skills and Claude.MDs for them.
MEGA retweeted
We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.
We'll begin restoring access tomorrow, and will share an update soon.
We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.
MEGA retweeted
tensors are live! huggingface.co/OBLITERATUS/G…
GGUFs currently uploading and should be up in 5-10 hours
