Просмотр новости
Найдите то, что Вас интересует
Спутники
РН
SpaceX
Роскосмос
NASA
СпутникиРНSpaceXРоскосмосNASA
Image
Amazon Web Services recently published a paper defining the AI-Driven Development Lifecycle, or AI-DLC — a reimagined SDLC (Software Development Lifecycle) for a world where AI, not humans, initiates the workflow. I read it the way I read most new methodology proposals by hunting for what's missing before I look at what's new. It didn't take long.To be fair, there's real thinking here. AI-DLC replaces Sprints with "Bolts" (yes, really), measured in hours rather than weeks. It folds design techniques such as Domain-Driven Design into the method rather than leaving them as homework. It reverses the conversation with AI, so AI proposes the plan and humans approve it, rather than humans prompting AI task by task. But a lifecycle isn't just the part between "I have an idea" and "the code shipped." And that's exactly the part AI-DLC focuses on at the expense of everything that makes software worth building in the first place. The question of value is missing from the lifecycle. Speed Isn't the Same as ValueAI-DLC measures itself in velocity. Bolts complete in hours. Mob Elaboration (a phase in the AI-DLC) condenses weeks into an afternoon. Deployment Units roll out to production faster than ever. All of that is impressive, and none of it answers the only question that actually matters - did we build something people needed?There's nothing representing a Product Goal in this method. No persistent expression of the outcome a Unit is meant to serve, and nothing resembling a Sprint Review, where real stakeholders look at a working increment and tell you whether you're still headed in the right direction. AI-DLC's events are internal. The mob validates AI's output; it doesn't step outside the room to check the output against the market. You can execute this entire method perfectly and still ship the wrong thing, faster.Where's the Retrospective?The paper doesn't just skip the Retrospective; it argues the team no longer needs one. Daily Scrums and Sprint Retrospectives, it says, were a byproduct of "longer iteration duration," and "proper application of AI leads to rapid cycles... rendering many of the traditional rituals less relevant." What replaces them? "Continuous, real-time validation and feedback mechanisms," a phrase that never gets defined as an actual technique anywhere in the method.Mob Elaboration and Mob Construction are phases for building. Neither one asks whether the way the team just built something actually worked or how we can improve it. A retrospective exists for that exact purpose: not to inspect the product, but to inspect the process and the team's own ability to deliver it. Dismissing that as a symptom of slow iteration, rather than a distinct discipline, is a mistake. Without some inspect-and-adapt loop on the process itself, "hours instead of weeks" just means a team can ossify into a bad workflow ten times faster than before. So, can a retrospective run continuously, as alluded to in the process? Experience would indicate that it is important to get the humans together, coupled with a bound problem and real data. That is how a good retrospective works. The bound problem is the Sprint Goal and the Increment. The real data is how the Sprint went. Having regular checkpoints encourages the team to come together regularly; if it runs continuously, from my experience, it never happens. There is always something more important! Collapsing Roles Isn't the Same as Building a TeamAI-DLC is candid about what it's doing to the org chart: infrastructure, front-end, back-end, DevOps, and security converge into a single developer role, with AI absorbing the specialized decision-making. Product Owners and developers remain "for oversight, validation, and strategic decision-making."Here's the part the paper skips. It tells developers what to validate. It doesn't tell them how to trust each other, how to disagree well when AI's recommendation and a teammate's instinct conflict, or how a junior engineer builds judgment when AI has already made most of the judgment calls for them. Psychological safety, skill development, and team accountability aren't footnotes to a delivery method; they're the reason a team can absorb this much change in the first place, and AI-DLC treats them as out of scope.How about starting with ScrumOf course, I could be considered biased from my role at Scrum.org. Actually, my experience with SDLCs began with the Rational Unified Process, where I served as the product manager. And I appreciate the nods to two RUP phases in the AI-DLC. Scrum provides a structure that allows teams to plan, execute, and review with a focus on goals. It encourages transparency via events and artifacts. It provides clear accountabilities, allowing focus and clarity of role. I think it provides a great structure for extending with the practices enabled by AI. In AI-SDLC parlance, AI can plan the Bolt, write the code, and run the tests in a fraction of the time it used to take. It still can't sit in a room with your customers and tell you whether you built the right thing. That's still on us. It can’t build a team that trusts itself to make judgment calls and choices. Humans are important parts of the lifecycle, and the SDLC must be built around them, not just the tools that enable the work!
| # | Наименование новости | Тональность | Информативность | Дата публикации |
|---|---|---|---|---|
| 1 | Scrum and AI: AI Doesn’t Replace Scrum, It Sharpens its Focus | 0 | 9.59 | 04-08-2026 |
| 2 | The New AI Operating Model. Start By Using Scrum | 0 | 6 | 07-07-2026 |
| 3 | Four Main Ways to Learn AI for Scrum Masters | 5 | 7 | 06-07-2026 |
| 4 | How to Avoid AI Debt by Borrowing from Agile Artifacts | 0 | 4.94 | 27-07-2026 |
| 5 | [Vlog] AI-native Scrum Team Specifications | 0 | 5 | 22-06-2026 |
| 6 | AI Tools Evaluation & Approval Framework for Scrum Teams | 0 | 10.27 | 10-08-2026 |
| 7 | AI won’t speed up software delivery — nothing has | 0 | 20.9 | 04-05-2026 |
| 8 | AI hasn’t shifted the bottleneck from coding to code review | 0 | 17.19 | 16-07-2026 |
| 9 | You Already Have an AI Working Agreement. Write It Down. | 0 | 6.14 | 12-07-2026 |
| 10 | Managing AI Risks in your Definition of Done | 0 | 8.24 | 27-07-2026 |