Question 1

What is technical debt?

Accepted Answer

The **future cost of a shortcut taken now** — the extra work created when code is written to be delivered rather than to be maintained. The metaphor is deliberate: like financial debt, it can be a sensible instrument, and like financial debt, it accrues interest.

**Where the term came from.** Ward Cunningham coined it to explain to non-technical stakeholders why a working product still needed further investment. His original meaning was narrower than current usage: shipping code reflecting your **current understanding**, then refactoring as understanding improves. Failing to do that second part is where the debt sits.

**The useful distinction** between kinds:

**Deliberate and prudent** — "we ship now and fix the architecture next quarter, and here is why". Legitimate, and frequently correct.

**Deliberate and reckless** — "we don't have time for design".

**Inadvertent and prudent** — "now we've built it, we can see how it should have been done". This is Cunningham's original case and is unavoidable.

**Inadvertent and reckless** — not knowing what good practice looks like.

**What the interest actually looks like.** Not a one-off cost, but a **permanent tax on every future change**: slower feature delivery, more bugs, changes that require touching many places, risky deployments, difficulty onboarding, and engineers avoiding parts of the system. **The symptom leadership notices is that everything takes longer than it used to**, with no single cause identifiable.

**Where it accumulates:** outdated dependencies; missing tests, which is the compounding kind because it makes every other repayment riskier; duplicated logic; undocumented decisions; dead code and unused feature flags; workarounds for problems since fixed; and infrastructure nobody dares touch.

**How it is actually repaid:** continuous small refactoring alongside feature work, which is far more effective than a rewrite; the "boy scout rule" of leaving code better than you found it; paying it down **where you are about to work**, since debt in stable untouched code costs nothing; tests first, to make change safe.

**Rewrites usually fail**, because they discard accumulated knowledge of edge cases while the old system continues to change.

Question 2

What is vendor lock-in, and how do you avoid it?

Accepted Answer

A situation where the cost of leaving a supplier exceeds the benefit of doing so — not because a contract forbids it, but because migration is expensive, risky or technically impractical. It is

Question 3

What is the difference between training, fine-tuning and retrieval?

Accepted Answer

Three different ways of getting a model to work with particular information, differing enormously in cost, and frequently confused when people ask about "training it on our data". Pre-training.

Question 4

Why do old files become unreadable, and what is digital preservation?

Accepted Answer

Because a file is only meaningful to software that understands its format, and both the software and the medium it sits on decay — on timescales far shorter than people assume. The three separate

Question 5

What is a zero-day vulnerability?

Accepted Answer

A zero-day is a security flaw that the vendor does not yet know about, or has known about for zero days — so no patch exists and defenders have had no opportunity to prepare. The related terms are

Question 6

What is synthetic data, and why is it used?

Accepted Answer

Data that is generated rather than collected — created by simulation, by statistical modelling of a real dataset, or increasingly by a machine learning model — and used where real data is

Question 7

What is the difference between proprietary, permissive and copyleft licences?

Accepted Answer

What you are permitted to do with the software, and — critically for the last of them — what you must do if you distribute your own version. Proprietary. The copyright holder grants a limited licence

Related Questions