Skip to content
SDLC in the AI Era

About & method

How this guide is made, and how far to trust it

A dated snapshot of how 12 software roles are changing with AI, researched with AI assistance and checked claim by claim against its sources. It is not peer reviewed.

What it is, and who made it

Pick a role and the site shows how the work changes, the risks, skills and tools, and ready-made AI instructions.

  • 12 roles
  • 84 described changes
  • 45 tools
  • 36 starter AI instructions

It is written by an independent editor and built with Claude Code. Claude Code is an Anthropic product, and a few sources on the site are published by Anthropic; they are labelled vendor-affiliated like any other.

How content is made

  1. 1

    Research

    An AI research agent reads current sources for one role and writes the page: each change, risk and tool with the sources it relies on.

  2. 2

    Link check

    A script opens every external link in the content, and any link that doesn't resolve is fixed before the page goes out.

  3. 3

    Independent fact-check

    A separate agent, which cannot edit files, compares every claim with its cited page and reports anything the source doesn't say.

  4. 4

    Fix and re-check

    The editor fixes each reported problem, and the page is checked again until the report is clean.

Rules every page follows

  • Every described change cites at least one source.
  • Statistics appear only when the source contains that exact figure, and the passage is quoted in the source list.
  • Claims about the AI era use sources published in 2024 or later. Older sources describe only how work used to be done.
  • A finding from one company, study or person stays attributed to them ("At Google…", "In one study…"), and caveats such as small samples are kept.
  • No vendor hype, and no "latest" or "new" without a date.

What “Our suggestion” means

Some sections are marked Our suggestion: first steps, which tools to start with, related roles and the starter AI instructions. They are the editor's practical advice, based on the sources, and are not sourced claims themselves.

Review dates

Each role page shows “Last reviewed” with the date its content was last researched and fact-checked, so dates can differ between roles. Tools, prices and reports change quickly, so treat anything older than a few months as a starting point and check the linked source.

Sources and how they are labelled

Role pages cite 95 sources in total (93 distinct pages; a few are cited on more than one role). Each one is labelled by what the cited page is, and whether its publisher sells products or services in that space.

LabelMeaningSources
Research paperAn academic paper or preprint (arXiv, conference, journal).28
SurveyA report built on questionnaire or interview responses from many people.16
Official guidanceGuidance from a standards body, professional association, government, framework authors, or an open-source project documenting its own standard.6
Vendor pageA company's documentation, blog or case study about its own product, customers or internal practice.20
PractitionerAn individual expert's or consultancy's article, newsletter, blog or technology radar.22
NewsJournalism or news analysis.3

44 of 95 are labelled vendor-affiliated: the publisher sells in the topic's space, as with a tool company's own survey or a consultancy's radar. That doesn't make a source wrong, but it's a reason to read it with its incentives in mind. When the label was unclear, the more cautious one was chosen. The evidence badge on each role page counts every source not labelled vendor-affiliated as independent.

Every dated source was published in 2024 or later. 13 sources show no publication date on the page, such as product documentation.

Vendor-affiliated sources by role

Roles differ in how many of their sources come from publishers who sell in that space. Where the share is high, give the source links a closer look.

Known limitations

  • AI-assisted research and checking reduce mistakes but don't remove them. Follow the source link before relying on a claim.
  • Only English-language sources were used.
  • Many industry reports are published by companies with something to sell; the labels above show where.
  • Tools and reports change quickly, and the site is not updated continuously.
  • Roles vary a lot between companies. The pages describe common patterns, not how every team works.

Privacy

The published site counts page views and a few clicks, such as copying an AI instruction or opening a source, with GoatCounter, an open-source analytics service. According to GoatCounter's privacy policy (opens in new tab), it doesn't store IP addresses, the full browser User-Agent or any tracker ID, and it doesn't store anything in your browser with cookies or local storage. That is why this site has no cookie banner.

Found a mistake?

Open an issue on GitHub (opens in new tab) with the page, the claim and what the source actually says.