Do

Podcast 2 — In conversation

Speaker A: I was teaching English grammar to an advanced class last week and realised how bizarre do-support looks when you try to explain it to someone who doesn't already have it instinctively. Why do we need do to form questions and negatives at all?

Speaker B: It's genuinely one of the most interesting puzzles in English grammar history. Most European languages manage perfectly well without an auxiliary for questions and negatives — French says je ne sais pas, German says ich weiß nicht. English developed do-support in the late medieval period, and linguists still debate exactly why. The best explanation seems to be that as English lost its inflectional verb endings — the endings that marked tense and person — the do auxiliary filled the structural slot those endings had occupied, providing a carrier for tense and negation when the main verb had moved.

Speaker A: The emphasis use of do is fascinating — I do mean it, she does care, I did warn you. It adds a kind of extra insistence that's hard to replicate without it.

Speaker B: The emphatic do is one of the most expressive tools in English — it signals that the speaker is pushing back against an implied denial or doubt. I do mean it implies someone questioned whether you meant it. She does care implies someone suggested she didn't. It's what linguists call a contrastive focus marker — it puts the entire weight of assertion behind the verb. No other structure in English does quite the same work, and the moment you try to express the same insistence without it — I truly mean it, I genuinely care — it takes more words and often sounds slightly less natural.

Speaker A: Close synonyms for the main verb sense — perform, execute, carry out, accomplish, achieve. How does do compare?

Speaker B: Do is the most general and least informative of the group — it simply means that an action was performed. Perform implies a degree of skill or presentation — you perform a task or perform for an audience. Execute implies precision and following a procedure. Carry out implies following through on a plan or instruction. Accomplish and achieve both imply successfully reaching a goal. Do is the unmarked default — you do things when you don't need to specify how, with what skill, or to what standard. The more specific synonyms add dimensions that do deliberately leaves out.

Speaker A: Common mistake — using do when a more specific verb would be clearer. Do a meeting, do a call, do a document are vaguer than hold a meeting, make a call, write a document. Do as a substitute for all verbs is a habit worth noticing in professional communication.

Speaker B: Business English in particular has developed a habit of using do as a catch-all main verb that strips away useful semantic information. Do a review tells you much less than conduct a review, which tells you much less than critically evaluate the assumptions underlying our strategic review. The more precise verb, the more useful the communication.

Podcast 3 — Prompt engineering

Speaker B: Do is a deceptively powerful word in AI prompts — it's the default action verb that AI systems use when no more specific instruction is given, and replacing it with precise verbs dramatically improves output quality. Let's explore this.

Speaker A: This is actually one of the most practical things I've learned about prompt engineering — that vague do instructions produce vague outputs, and replacing do with a specific verb transforms the result.

Speaker B: First prompt — and notice how we change what do means by replacing it.

Refactor this Python function to improve its readability: strip do-nothing comments, rename variables to reveal their purpose, extract any logic block longer than 10 lines into a named helper function, and replace any loop that can be expressed as a list comprehension. Do not change the function's observable behaviour — only its internal structure. Show a before-and-after comparison with a brief explanation of each change made.

Speaker B: Do not change the observable behaviour — here do is used in a constraint: the refactoring must not do what refactoring often inadvertently does — break the external contract of the function.

Speaker A: What about using do as an AI task instruction framing?

You are a senior code reviewer. Do a thorough review of the following TypeScript module and produce findings in three categories: critical issues — bugs or security vulnerabilities that must be fixed before deployment, quality issues — things that do not break functionality but reduce maintainability, and style issues — things that do not affect quality but deviate from the project's coding conventions. For each finding, state what was done wrong, what should be done instead, and why.

Speaker B: What was done wrong, what should be done instead, and why — the review prompt uses do as a past tense description of the problem and a present tense prescription for the fix.

Speaker A: What about replacing do with specific action verbs to improve prompt precision?

Transform the following vague AI prompt into a precise one by replacing every instance of do with a specific action verb. For example: do the analysis becomes critically evaluate the assumptions in the dataset, do a summary becomes synthesise the three most significant findings into a single paragraph, and do something about performance becomes identify the two highest-impact optimisations and implement the one with the lowest regression risk. Show the original and transformed prompts side by side.

Speaker B: Replace every do with a specific action verb — the prompt engineering tool uses do-replacement as a systematic technique for improving prompt precision.

Speaker A: What about a coding task with do as orchestration?

Write a Node.js task runner that interprets a plain-English do instruction and dispatches it to the appropriate handler. For instructions containing the verb do followed by a noun phrase, parse the noun phrase to determine task category — do a backup maps to the backup handler, do a health check maps to the monitoring handler, do a deployment maps to the CI-CD handler. For ambiguous do instructions, return the top three interpretations ranked by confidence and ask for clarification.

Speaker B: Top three interpretations for ambiguous do instructions — the task runner models the exact ambiguity that makes do such a challenging word in command interfaces.

Speaker A: What about a testing tool?

Write a Python test suite generator that creates comprehensive test cases for a function based on what the function is supposed to do. Accept a function signature and a natural language description of what it should do. Generate: positive tests — inputs where the function should do what it says — boundary tests — inputs at the edges of the expected range — negative tests — inputs where the function should not do what it would do with valid input — and error tests — inputs that should cause the function to raise specific exceptions. Output a pytest test module.

Speaker B: What the function should not do with invalid input — the test generator uses do as the semantic anchor for specifying correct behaviour, making what the function should and should not do the dual frame for test coverage.

Speaker A: Do in prompts is most powerful when it anchors the action frame clearly — but replacing vague do instructions with precise action verbs consistently produces more useful, specific, and actionable outputs.

Speaker B: And the single most effective prompt engineering habit you can develop is to look at every do in your prompt and ask: what exactly do I mean — and can I say that more precisely?

Speaker B: Prompt example: