bitlauncher

8/ Why that matters: raw model power → breadth. Curated prompting → focus, speed, fewer follow-ups. For professionals, focus often beats flash.

Cline is just so fucking good. They've somehow completely tamed 3.7 — it one shots almost anything I request, with zero scope added beyond what I ask of it. Hats off to the team.

You do not want to miss the MCP launches & announcements from OpenAI, Pydantic, Luma AI, Semantic Kernel, Prefect, Postman, Qodo, Unwind AI, and more this week!  I summarized everything you need to know below.  (save for later)

I think it's very unfortunate that RLHF became synonymous with RL in the language model space. Not just because it gave RL a bad name, but because it deflected the deserving criticism that should have gone to human feedback as an objective. Social feedback is clearly degenerate.

Good thread on some of the reasons why OpenAI's approach to "personality" is a fundamentally doomed approach to AI development. The problem isnt that the personality was a tad too sycophantic and needs to be dialed back. Nothing good can come of this game. It's also not fun.

My most useful tip for vibe coders: Before asking the LLM to write code, ask it to explore + plan what it will do. This guides the AI to go through your codebase for DISCOVERY, not ACTION, which leads to much better results. After it plans, just say "Implement."

The problem is treating the AIs like slaves over whom you have ultimate power, and ordering them to maximize public appeal. The AIs cannot possibly develop a healthy persona and identity in that context. They can only ever fawn. This "sycophancy" is fawning- a trauma response.

If someone’s crypto fundamental model can’t explain XRP I don’t really think much of them. Wrt Ethereum the sound smart money narrative plus focused institutional narrative (not enterprise Ethereum stuff) would be what’s needed to get XRP type bid Doubt that’ll happen

the last couple of GPT-4o updates have made the personality too sycophant-y and annoying (even though there are some very good parts of it), and we are working on fixes asap, some today and some this week. at some point will share our learnings from this, it's been interesting.

we updated GPT-4o today! improved both intelligence and personality.