You cannot predict which instructions a model needs, so a system prompt has to be discovered by running the thing rather than designed in advance.

Added to korrents 27 Jul 2026 · How quotes work · Something wrong? Tell us

Related korrents

Our reading — they may agree, disagree or merely touch the same thing. Closest first: a shared subject counts for most, then how near the wording is.

• A frontier model is measurably more intelligent with no system prompt at all; the prompts that remain are there for the product, not the model. BCBoris Cherny27 Jul 2026similar wording
• Predictions that nobody will write code soon dodge the question of what code is: prompting a model is still producing something we will have to read and change. MFMartin Fowler7 Apr 2026similar wording
• Prompt injection is unsolved, but the latest models are post-trained to resist it, so it takes far more than 'ignore all previous instructions' now. PSPeter Steinberger12 Feb 2026similar wording
• Context has two budgets, not one: the information budget everyone thinks about, and an instruction budget, where conflicting instructions cost the model real computation to ignore. DHDex Horthy15 Jul 2026similar wording
• Reasoning will generalize the way instruction tuning did: add enough verifiable domains and, at some point nobody can yet locate, the rest start working on their own. NLNathan Lambert3 Feb 2025similar wording

On the map

Open the map on its own page →