Every term these docs use, defined once, with a link to the page that covers it.
A few words in these docs mean something specific, and several of them sound like each other. This page defines each one once and points at the page that covers it properly.
How work is organised
Term
What it means here
Chat
One thread of work, listed in the sidebar, with a title and everything produced inside it. See Chats, turns and runs.
Turn
One exchange inside a chat: what you asked, and what came back.
Run
The work a turn dispatched. This is the unit credit is spent in, and the thing that has a step indicator and a ceiling.
Follow-up
Any message after the first one in a chat. It starts from what the previous run produced rather than from nothing.
Workspace
The files a run works in. Either empty (from scratch) or the project you uploaded; see Workspaces.
What a run is asked to do
Term
What it means here
Intent
Which of the three jobs a turn is: Build, Debug or Optimize. It belongs to the turn, not the chat; see Build, Debug and Optimize.
The seven switches deciding what a run may read and how hard it checks itself. They apply to Plan-mode Build runs; see Tools overview.
Reasoning effort
How long the model may think before committing to an answer. Separate from the effort setting on an optimize run; see Reasoning effort.
Run defaults
What every new chat starts from, saved in Settings under Run defaults. Per-run changes in Run setup do not overwrite them.
What comes back
Term
What it means here
Artifact
Any file a run produced: the module, the testbench, the simulation log, the synthesis report.
Change report
What a project run returns instead of a single module: added, modified and deleted files with a diff for each. See Reading the change report.
Proposal
A change Debug or Optimize would make, shown as a diff and applied only if you say so. See Applying a proposed fix.
Stale proposal
A proposal built against files that have since changed. It is not applied; ask again rather than merging by hand.
Verdict
A stated outcome rather than a guess: the simulation result, the diagnosis, or the behaviour check on a rewrite. Each one says how strong a claim it is making.
Coverage summary
The evidence strip on a thorough verification run: scenarios, checks, random vectors and mismatches. See Exhaustive Verification.
Credit and plans
Term
What it means here
Plan credit
The balance a plan gives you. Refills monthly, spent first, does not roll over. See Credit and top-ups.
Usage credits
The balance you buy separately. Never expires, and takes over when plan credit runs out.
Welcome grant
The one-time credit on a new account. It never expires, and it is held in reserve while a paid plan is running.
Free runs
A referral reward, counted in runs rather than money and spent before either balance. See Referrals.
Reservation
Credit set aside when a run starts and returned when it settles, so two runs cannot spend the same last dollar.
Credit pack
A one-off purchase that adds usage credits without changing your plan.
Hardware words we use in a particular way
Term
What it means here
Self-checking testbench
A testbench that reports pass or fail itself, rather than producing waveforms for you to read.
Synthesis estimate
A resource figure from our own synthesis pass, not from Vivado or Quartus. Good for comparison, not for a datasheet; see Reading the numbers.
Logic depth
The longest chain of logic between registers. A proxy for timing pressure, and not a timing report.
Bounded check
A behaviour check that holds for a stated number of cycles rather than for all time. See Equivalence checking.
Divergence
A construct that behaves one way in simulation and another on a device. See Static checks.