Decoder. plain-English AI glossary

SWE Agent

▲ Rising

An agent built to resolve real software issues end to end — read the repo, write the fix, prove it works.

Think of it like

Handing a junior engineer a bug ticket and getting back a tested, ready-to-merge pull request.

Example

Given a GitHub issue, the agent locates the buggy files, edits them, runs the test suite, and produces a passing patch.

How it actually works

SWE agents are the benchmark-driven face of agentic coding, measured on suites like SWE-bench where the task is “given this repo and this issue, produce a patch that passes the hidden tests.” Progress here has been fast because the scoring is objective. The gap to everyday work is that real tickets are underspecified and touch design, review, and taste — things a hidden test can’t capture.

For product teams

The clearest yardstick for “can AI actually do engineering work,” with a real-world asterisk.

For engineers

An agent evaluated on repository-level issue resolution, producing patches validated against held-out tests.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome