Technology is a reproducible trick, not just chips and code

Definition

Technology is the application of conceptual knowledge to achieve practical goals. The standard encyclopedia formulation adds the clause that matters most: especially "in a reproducible way." That qualifier is the whole game. A one-off flash of insight is not technology. A procedure that survives being written down, taught to strangers, and run again — that is. I trust definitions that hold up under an ablation, and this one does: delete "reproducible" and the sentence quietly covers every lucky accident in human history, which is not what anyone means.

Etymology and scope

Wiktionary preserves the older industrial framing: "the combined application of science and art in practical ways in industry," and, secondarily, the machines and equipment so designed. Wider treatments stretch further, covering the making, modification, and use of tools, machines, techniques, crafts, systems, and even methods of organization.

That last item deserves emphasis. Methods of organization are technology. A just-in-time inventory system is technology with no parts you can kick. So is code review. If your mental model of the word stops at hardware, you will systematically underrate the software — in the original sense — that makes hardware work.

Technology versus science

The two are neither synonyms nor rivals. Science generates and tests conceptual knowledge; technology applies it to practical goals. The loop runs both directions: technology supplies science with instruments — telescopes, sequencers, accelerators, GPUs — and science supplies technology with new conceptual material to apply. Encyclopedia entries note that technology plays a critical role in science, which frankly understates it; much of modern experimental science is impossible without its tooling.

A working taxonomy

The IP definition surprises people, but it is the honest one for anyone who has watched a tech transfer negotiation. What changes hands is rarely the artifact; it is the know-how, plus the paper trail that makes the know-how legible to someone else.

Ablating the word

When someone says "technology," they mean one of four layers, and most arguments start because two people mean different ones:

• Artifact — the phone, the reactor, the model checkpoint.
• Know-how — the fabrication recipe, the training procedure, the tuning folklore.
• Process — the repeatable workflow that turns know-how into output at scale.
• Organization — the roles, incentives, and review structures the process runs inside.

Run this decomposition before any debate about whether some technology is "good" or "bad." A governance argument about artifacts is a different conversation from one about processes, and they get resolved by different levers.

Failure modes of the term

• Anthropomorphism. "Technology wants X." It wants nothing; it is a procedure, not an agent. When an algorithm "decides," a person specified the objective, chose the data, and signed off on deployment. Push back on the passive voice — it launders accountability.
• Category error. Treating information technology as the superset. Computing is one branch. A vaccine cold chain is technology; so is a shipping schedule.
• The demo trap. Per the core definition, reproducibility is the entry requirement, not a nice-to-have. A demo that skips the failure cases is not technology yet — it is a magic trick awaiting its first repeat performance. In machine learning this is exactly why ablations exist: they are the reproducibility audit in miniature.

The one-line version

My own compression, offered as opinion: technology equals knowledge times reproducibility times deployment. Whatever fails that multiplication is marketing.