How I see software dev in AI era
The article proposes a model of software development as nested optimization, where an AI agent refines code against tests while a developer refines the agent's understanding of the software's intent. It defines test suites as predictors of production failure, analyzing their soundness and completeness to measure the gap between tested behavior and actual intent. The text concludes that this optimization target is constantly shifting because external pressures continuously update the software's intent before the previous version fully converges.
- ▪The model describes software development as a process where an AI agent refines code against failures while a developer refines the agent's understanding of the software's intent.
- ▪Test suites are evaluated based on soundness, which ensures test failures represent real production issues, and completeness, which ensures all production failures are detected.
- ▪The distance between a test suite and production errors is defined as the probability that code fails in production while the test suite remains green.
- ▪External factors such as sales requests and market changes create an update pressure that continuously shifts the software's intent, preventing final convergence.
Hacker News (AI / LLM) files mainly under ai. We currently carry 5,645 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | dhilst |
| Canonical URL | https://dhilst.github.io/2026/09/19/sweng-as-opt-prob/ |
| Publication time | Sat, 19 Sep 2026 19:44:56 +0000 |
| Retrieval time | 2026-09-19T19:58:46.099Z |
| Last seen | 2026-09-19T19:58:46.099Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | D3YL7bg4hd7j · 1 stories |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
How I see Software Development in the AI era 19 Sep 2026 Most software is not proven correct before release. It is tested, judged good enough and shipped. When it fails in production, we fix it, add a test and ship again. This post proposes a model of that process: software development as nested optimization. Inside, an AI agent refines code against failures and tests. Outside, a developer refines the agent’s understanding, tests and fixes against the software’s intent. The model Software has an intent: what users, developers and the business expect it to do. Intent is larger than any written specification. Some of it becomes visible only when the software violates it.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at dhilst.
Discussion
0 commentsMore from dhilst
![AI Actress interviewed by Piers Morgan [video]](/img-proxy?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fx4yyBWkYSHs%2Fmaxresdefault.jpg&w=320)
