How Claude clusters feedback into synthetic evals
People ask ChatGPT about your category. AI names a few brands. Is yours one of them?
SEO is dead. ChatGPT, Gemini, and Perplexity give one answer naming three brands — and those brands are shaped by Reddit, the #1 source LLMs trust. If yours isn't in those threads, you don't exist.
ReddGrow finds the discussions AI already cites for your keywords. You join the conversation. When AI re-indexes, you're part of the answer. Four steps from invisible to recommended.
Today's top 10 insights for PM Builders from X and LinkedIn.
How Claude clusters feedback into synthetic evals
#1 đť•Ź
Peter Yang breaks down Anthropic’s build of the next Claude with Alex Albert: they co-design the model and harness, use Claude to cluster user feedback into synthetic evals, and train its character and personality.
#2 in
Marc Baselga shares Sebastien Goddijn’s insight that Ramp’s AI adoption only drove real value after engineers built context files, MCPs, memory and workflows. Without this scaffolding, non-technical staff using Claude, ChatGPT or Cursor foot the hidden “setup tax.”
#3 đť•Ź
Santiago shows that pressing CTRL+G in Claude Code opens a full-featured editor for prompt writing, making long prompts 100Ă— more manageable than typing them directly in the terminal.
#4 đť•Ź
Teresa Torres highlights that Rhea’s Factory uses AI to optimize the entire enzyme production process—focusing on cost-driving parameters rather than just boosting enzyme performance. This end-to-end approach delivers scalable, sustainable, and low-cost products.
#5 đť•Ź
Dharmesh Shah argues that legacy APIs assumed human developers who’d read docs and iterate, but as agents become the primary users, APIs, MCPs, and CLIs must be redesigned to be more discoverable, legible, and forgiving.
#6 đť•Ź
Garry Tan announces that GBrain now ships with ZeroEntropy as its recommended default embedding and re-ranking engine, replacing OpenAI and Voyage AI.
#7 đť•Ź
Yann LeCun argues that current LLMs underperform on continuous, high-dimensional, noisy data, revealing a critical blind spot in their processing capabilities.
#8 đť•Ź
Dharmesh Shah applauds HubSpot for topping @jasonlk’s “agent readiness” list, underscoring that software must deliver not only stellar human UX but also robust agentic experiences (AX).
#9 đť•Ź
Garry Tan observes that we’ve shifted from writing code to invoke LLMs to authoring prompts and skill files that let them execute code—and hints that the next phase of this evolution is still unwritten.
#10 in
Peter Yang recaps Alex’s talk on prepping AI products for next-gen models. Key tips include testing on the latest model, pruning outdated prompts like bonsai, and trusting the AI to handle its own reasoning.