Scrum is obsolete, and the strongest evidence is not an opinion, it is the job market. Scrum Master roles are being merged into Engineering Manager roles or removed entirely. Job descriptions are renaming themselves. Hiring is the most honest signal you get about whether a framework is still load-bearing, and the signal says no. Alex Polyakov's Analyst's Corner piece makes the case well. Where it stops short is naming what replaces Scrum, which is the move that lets every Scrum org agree with every paragraph and keep doing Scrum on Monday.
Why it is obsolete
A few things are true and worth repeating. Scrum borrowed Extreme Programming's iteration cadence and dropped the engineering rigor. XP required pair programming, test-driven development, continuous integration, and refactoring. Scrum shipped the cadence without those practices because they were hard to sell to non-engineering buyers. The trade-off worked for adoption and failed for rigor.
Scrum also normalizes failure. When everything is an experiment, nothing is truly late. The retrospective converts missed commitments into learning opportunities indefinitely, and when learning is the success criterion, delivery becomes optional. And ownership dissolves into consensus: decisions belong to the team, which means they belong to nobody in particular. Newer engineers see the theater first and develop a fast nose for when planning is performance rather than planning. That instinct is correct.
The named replacement
A four-to-six person builder pod. One PM, one designer who prototypes in code, two to three engineers, owning one customer-facing outcome end-to-end. Small enough that everyone holds the work in their head, large enough to ship in parallel streams.
Three artifacts, total. A working prototype, built in Claude Code or Cursor in four to six hours per iteration and real enough for customers to use. A five-row eval suite that defines done as falsifiable behavior, not a checklist. And an outcome ledger with three to seven active bets, each with a decision deadline two to four weeks out. There is no backlog, no user stories, no velocity, no story points, no sprint commitment.
The cadence is weekly: a 30-minute review of each active bet, a five-minute Friday update on what moved, and a quarterly retro that is a real retro. The roles that do not exist: Scrum Master, Product Owner as a separate role from PM, Release Train Engineer, Agile Coach. None of them have work to do here. I have run this structure with several teams over the last 18 months, and the numbers, time from signal to shipped prototype, eval pass rate at first ship, decisions per quarter, outperform every Scrum team I have measured against, with far less ceremony.
Agile the philosophy survives
The agile manifesto's four values are defensible. Individuals over processes, working software over documentation, collaboration over contract negotiation, responding to change over following a plan. None of them require Scrum, sprints, or story points. The builder pod is more aligned with the manifesto than Scrum has been in years.
If you run Scrum and cannot kill it yet, retire the sprint commitment this week. Keep the cadence if you need the rhythm, drop the commitment language, and replace it with the outcome ledger: three to seven bets, each with a decision deadline. Most of the ceremony tax falls out without a re-org.