HM

Podcast 2 — In conversation

Speaker A: I sent a proposal to my manager and the only thing she said in reply was hm. Just hm. And I spent the rest of the day trying to decode it.

Speaker B: That is the essential mystery of hm. In spoken conversation you have prosody — the rise or fall of the voice, duration, facial expression — to decode it. In writing, hm is deliberately ambiguous. It could mean interesting, I am thinking about it, or I am not sure about this. Your manager has left maximum interpretive space.

Speaker A: Is there a way to make it clearer in writing? Like does hmm mean something different to hm?

Speaker B: Conventionally, yes. Hm tends to be a brief cognitive acknowledgement — I heard you, I am processing. Hmm is longer, more genuinely thoughtful. Hmmm suggests deeper uncertainty or even mild concern. And hmmmm with several m letters can be playful or theatrical depending on context. The extra letters signal duration of the hum in the speaker's head, so more m letters generally means more time spent thinking.

Speaker A: What about in dialogue writing? I always wonder how much hm to use.

Speaker B: Used sparingly, hm is one of the most effective tools in a dialogue writer's kit. It breaks up the rhythm, signals character, and creates natural pauses that make dialogue feel human rather than scripted. A character who says hm before answering a difficult question feels more real than one who launches immediately into a polished response. But overuse flattens it — it becomes a tic rather than a meaningful signal.

Speaker A: Is it ever appropriate in formal writing?

Speaker B: Almost never in formal documents, reports, or professional correspondence — it would undermine the authority of the writing. But in scripts, transcripts, informal journalism, creative non-fiction, and personal essays, hm can appear naturally. And in dialogue within fiction, it is entirely at home. The register is irreducibly informal and spoken.

Speaker A: So hm is a word that belongs to the spoken world and only partially translates to the written one.

Speaker B: That is a beautiful way to put it. Hm is native to the mouth, not to the page. Every time it appears in writing it is an attempt to import the sound and rhythm of real human thought into text — and when it works, it is one of the most economical signals of authenticity a writer has.

Podcast 3 — Prompt engineering

Speaker B: Today we are using hm as a thinking and ambiguity keyword in developer prompts — for generating systems that model uncertainty, represent thoughtful states, capture hesitation in user experience flows, or build interfaces that acknowledge when a decision is not yet made.

Speaker A: Hm as an ambiguity and deliberation signal. In prompt engineering it tells the AI you want to model the in-between state — not yes, not no, but still thinking. That is a surprisingly rich design concept.

Speaker B: Exactly. Hm as a state in a system is the honest acknowledgement that not everything is binary. Let me show you.

Speaker B: Design a decision workflow component for a project management dashboard that includes an explicit hm state alongside approved and rejected, allowing team members to mark a proposal as hm when they have read it but are still deliberating, displaying hm items in an amber pending-thought queue separate from the main approval pipeline, and sending a gentle reminder after forty-eight hours if a proposal remains in hm state.

Speaker A: Hm as a named workflow state. The amber pending-thought queue makes ambiguity visible and manageable without forcing a premature decision. That is excellent UX for complex approval processes where real thinking takes time.

Speaker B: Right. For a database schema.

Speaker B: Create a database schema for a content review platform where each content item can have a status of published, rejected, or hm, the hm status records the reviewer ID, a free-text deliberation note, and a review deadline, an index on status and review deadline enables efficient retrieval of overdue hm items, and a daily job escalates hm items past their deadline to a senior reviewer queue.

Speaker A: Hm as a first-class status in the schema with its own fields, index, and escalation logic. The deliberation note field is inspired — it is a place to park the thinking that makes the status meaningful rather than just a limbo state.

Speaker B: Good. For a UI prompt.

Speaker B: Build a feedback collection widget for a SaaS product that offers three response options: yes, no, and hm — where hm represents not sure yet, displays a small tooltip explaining that hm means I need more time or information, stores hm responses separately from yes and no in the analytics database, and generates a weekly insights report showing what percentage of users are in hm state for each feature, treating high hm rates as a signal of unclear value proposition.

Speaker A: Hm as a data point in analytics. A high hm rate on a feature is a meaningful signal — it tells the product team that users have noticed the feature but have not yet formed a clear opinion. That is far more actionable than grouping hm with no.

Speaker B: For a full application.

Speaker B: Build a full HR performance review platform where managers can score employee goals as met, not met, or hm for goals that are partially achieved or still in progress, the hm category stores a structured note with evidence, context, and a follow-up date, a hm dashboard surfaces all in-progress assessments approaching their follow-up date, managers receive automated prompts to resolve hm items before the review cycle closes, and the final review report shows what percentage of goals passed through a hm phase before resolution.

Speaker A: A complete performance review platform where hm is a structured, time-limited, evidence-bearing state. The follow-up date and automated resolution prompts turn thoughtfulness into a managed process rather than an indefinite limbo.

Speaker B: Hm in a prompt tells the AI you want to honour genuine ambiguity in your system design — to build space for thinking rather than forcing false binary decisions. It is one of the most human concepts you can embed into software, and naming it hm makes it immediately clear what state you are describing.