Standing service codebase-questions-answered-each-month-with-citations · revised 12 October 2026
Standing service
Ask up to 20 questions about your codebase each month, with file and line receipts
Each month up to 20 questions about your repository are answered with file and line citations at a named commit, checked by script. "Not in the code" is a valid answer. This is not a chatbot.
This starts a conversation by email. Nothing is charged, and nothing is read or run, until we have agreed scope, access and terms with you in writing.
The responsibility you hand over
New questions arrive every week while a team learns an unfamiliar codebase. A hosted chat tool answers fast but its answers cannot be checked, and asking a colleague costs their time.
Who it’s for: An agency, a new team or an acquirer that keeps meeting questions about a codebase they did not write.
Usually starts when: You have had a first set of questions answered and more keep coming.
The result: Each month up to 20 questions you send are answered in writing, each answer citing file and line at the latest commit and quoting the text. A script confirms every citation. Five trick questions a quarter are a spot check, not proof that we never invent an answer.
What stays true, and what we do about it
No response-time guarantee beyond the written turnaround is published for this new service.
Answers are written in working hours within a turnaround agreed in writing. There is no round-the-clock cover.
What must remain true
- Every answer cites a file and line that exists at the named commit and quotes the text
- An answer that the code does not support says "not in the code"
What we watch
- The repository at the latest commit, read-only
- The questions you send
When something happens
Scroll the table sideways to read it all.
| When | What we do |
|---|---|
| Questions arrive. | We answer them in the agreed working days with citations. Linked job: 25 codebase questions with citations |
| A question needs runtime or production data. | We say the code cannot show it and name what would. |
| A quarter ends. | We run the five trick questions your engineer wrote and report whether any got an invented answer. |
We do on our own
- Read the repository
- Answer questions with citations
- Say "not in the code"
We ask you first
- Wider access
- More than 20 questions in a month
- Any use of the answers outside your team
We escalate to you when
- A trick question gets an invented answer
- A citation still fails the script after two fixes
- The code can no longer be shared
How you know it held. Each month the citation log shows every file, line and quote matched, and each quarter your engineer rates 10 answers and at least 9 must be correct.
How we keep it true
This service is never finished. Each month's summary shows whether what it watches stayed true, and it continues until you end it.
Get 25 questions about your codebase answered, with file and line receipts Job Optional
Written answers to up to 25 questions about one repository, each citing file and line at a named commit, with "not in the code" where the code does not say. A script checks every citation.
Bought separately, once
The one-off set of 25 questions with 5 trick questions.
What is included, and what is not
- The monthly answers document
- The citation check log
- The quarterly trick-question results
Included
- One repository, read-only at the latest commit each month, up to 20 questions
- Each answer cites file and line, quotes the text and says "not in the code" where it is not
- A citation check log and a quarterly trick-question check
Not included
- A hosted chat assistant
- Opinions on quality or security
- Answers needing runtime or production data
- More than 20 questions a month
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
Each month a script confirms every citation exists and every quote matches at the named commit.
Evidence: The monthly citation log with zero mismatches.
Each quarter all five trick questions, written by your engineer and unseen by us, are answered "not in the code".
Evidence: The five questions and answers.
Each quarter your engineer rates 10 sampled answers (picked by script with a fixed seed), at least 9 are correct (the true rate is then at least 60% with 95% confidence), and none gives reasoning the quoted code does not support.
Evidence: The 10 answers and their ratings.
Sign-off. You read each month's answers and tell us whether the cap and turnaround suit you. A month counts as delivered when you have received the answers and the log.
If it fails. If the code cannot be read we say so and do not guess. If the service is not working for you, you can end it at the end of any month.
When it fits, and when we stop
It fits when
- A repository you may share read-only
- Questions written down each month
- A named engineer who rates a sample and writes five trick questions each quarter
We stop and tell you if
- The code can no longer be shared
- No questions arrive for two months
- The licence forbids sharing
What could go wrong
The service writes documents and changes nothing. Our copy of your code is deleted after each month's answers unless you agree otherwise.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| An answer invents reasoning around a real citation. | Quarterly trick questions and your engineer's sample rating test for it. |
| Answers reveal weak spots. | The document goes only to the person you name. |
| Code and comments carry instructions meant for the model. | Code and comments are treated as data: the model gets the text and returns answers a script checks, with no tools that change anything, no network and no secrets. |
A separate automated reviewer checks the work before hand-over. A named engineer on your side rates a sample each quarter. No human supervisor is included unless your agreement names one, and at launch the work is largely automated.
Questions
Is this a chat tool?
No. It is written answers with citations, in batches, each month.
Can we ask more than 20?
We agree a bigger allowance in writing, or batch the rest into the next month.
What if the code does not answer?
We write "not in the code" and say what would answer it.
Do I need the one-off set of 25 questions first?
No. The monthly service stands alone. The one-off set is for a first large batch of questions.
Also part of
Send an enquiry
Send us
- The language, rough size and the kind of questions
- Whether the code may leave your systems
- No code in the first enquiry
Later, once you agree
- Read-only code through a company-controlled route each month
- The questions, in writing
- Five trick questions each quarter, written by your engineer
- The name of your rating engineer
You keep the repository; we read the latest commit in a sandbox through a dedicated bot account with the least access that works (read-only first). Code and comments are treated as data: the model gets the text and returns answers a script checks, with no tools that change anything, no network and no secrets. Our automated agents read the files you send in a sandbox; we make no "no training" or "no storage" promise here, and the agreement names the model provider, country, retention and training status before anything is shared.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “codebase-questions-answered-each-month-with-citations” as the subject.