Sublime
An inspiration engine for ideas
When to use each SOTA AI model right now ✏️
↓ ↓ ↓
ChatGPT 4o - best ideas/invention but follows directions worst. I have screamed at the new 4o multiple times. Use this for first draft. Better at organization. GPTs still the best prompt efficiency hack. Stack them. Voice dictate whenever... See more
by deploying alternative news feed algorithms on in-silico social media platforms, where large language model (LLM) agents that mimic human social media users interact with one another, we can explore and test the impact of these alternative algorithms on macro-level social outcomes, such as conflicts and polarization.[385]
Audrey Tang • ⿻ 數位 Plurality: The Future of Collaborative Technology and Democracy
'Prompt Engineering for LLMs: The Art and Science of Building Large Language Model-Based Applications' - John Berryman & Albert Ziegler (Dec. 2024, PDF): https://t.co/hvgv4xLjKZ
Both were founding engineers for GitHub Copilot.
Stephen Reid
@stephenreid
Google Is Forcing OpenAI To Change Plans
OpenAI will change plans and release o3 and o4-mini instead of shipping GPT-5, which was supposed to combine all their models.
The big reason is Gemini 2.5 Pro, a thinking model that is very good at benchmark maxing (being #1 on... See more
Bindu Reddyx.com
ok so let me explain why subagents kill long context
Like you can spend $500m building 100 million context models, and they would be 1) slow, 2) expensive to use, 3) have huge context rot. O(n) is the lower bound.
Cog's approach is something you learn in day 1 of @CS50 - divide and... See more
swyxx.com