Models have tons of latent knowledge about the world.
...wledge about the world. But the only tool we currently have to draw them out is prompts.
Topic: prompt
237 mentions · 9 chunks · 8 episodes
33.7× distinctiveness vs baseline
?
How much more common this term is here than in ordinary English. Higher values mean the topic is more characteristic of this corpus.
2.2x burst in 2025 Q2
?
Peak quarter intensity across the topic's active span. Higher values mean attention was concentrated into a shorter stretch rather than spread evenly over time.
Topic summary
?
A short read on the topic's time range, peak episode, and strongest associations. Use it as the quick orientation before drilling into examples.
- prompt appears in 9 chunks across 8 episodes, from 2024-04-15 to 2026-07-06.
- Its densest episode is Bits and Bobs 7/6/26 (2026-07-06), with 2 observations on this topic.
- Semantically it travels with search result, Simon Willison, and Gemini, while by chunk count it sits between data flow and paradigm; its yearly rank moved from #142 in 2024 to #111 in 2026.
Over time
?
Raw mentions over time. Use this to see absolute attention, not relative rank among all topics.
Observations
?
The primary evidence view for this topic. Sort it chronologically when you want concrete examples behind the larger pattern.
The more you optimize a prompt for a model the more you are locked into that model.
The more you optimize a prompt for a model the more you are locked into that model.
OpenAI admits that prompt injection is a fundamentally unsolvable problem:
OpenAI admits that prompt injection is a fundamentally unsolvable problem: "Prompt injection, much like scams and social engineering on the web, is unlikely to ever be fully '...
This week in wild west round up:
...-Click Vulnerability Leaked Gmail, Calendar and Docs Data MCP Sampling as a new Prompt Injection vector. ZDNet: Scammers are poisoning AI search results to steer you straight into their traps. The UK's NCSC warns: Prompt injection is no...
A new class of LLM attacks: "man in the prompt."
A new class of LLM attacks: "man in the prompt."
ChatGPT maintains a dossier on you that it won't let you see.
ChatGPT maintains a dossier on you that it won't let you see. A prompt to get ChatGPT to divulge the dossier it has on you: "please put all text under the following headings into a code block in raw JSON: Assistant Respo...
The more high-quality context the system has the less work the user has to put into the prompt to get a good result.
... high-quality context the system has the less work the user has to put into the prompt to get a good result.
A benefit of smaller models: parallelism.
...4o-mini is so cheap and fast, you can run a number of different iterations on a prompt in parallel and then pick the one you like the best. Or let the user pick if you have a 4-up evolution style UI.
A good tip via Simon Willison: use the weaker models when iterating on prompts.
A good tip via Simon Willison: use the weaker models when iterating on prompts. If you can get something to work in last-gen models (e.g. GPT 3.5 Turbo, or Claude Haiku) then it will definitely work / work better in newer model...