Thesis

Once producing stops being scarce, a product role is defined by what it can judge, build or guarantee

Info

Originally written in French. Translated by AI — the meaning has been preserved, not the prose.

Angle

In a team where a developer interviews customers and a Product Manager pushes a pull request, the job title stops predicting anything: what situates a person is the rung they reach between the mockup and the production deliverable, and the kind of decision they answer for. The blurring of roles does not dissolve responsibilities, it makes them more costly to leave implicit — a product choice, a technical choice and an experience choice each keep their guarantor, and it is precisely because those three lines are held that everyone can step in beyond their home territory. What governs that opening is neither the discipline nor the title, but the risk of the area touched: a translation, a read-only screen and a billing rule do not call for the same standard, including in a regulated product. The one point that is redistributed to nobody and to no machine is commitment: someone has to stand in front of the others, six months later, and answer for the bet.

Synthesis

Taken one by one, these ideas look like field observations about a well-equipped team. Placed end to end, they describe how an organization can let its perimeters deform without losing track of who answers for what.

The starting point is an imbalance. Producing becomes much cheaper; deciding does not follow. If everything keeps coming back to the same place, the gain in production turns into a queue in front of the Product Manager, and the same scene replays step by step along the chain — the code speeds up, the product blocks, then the design, then the launch. It is that imbalance, and not a doctrine about roles, that pushes towards redistribution.

The redistribution happens in both directions, and both movements have their condition. Developers move up towards the product because delegating implementation to an agent forces them to write down what they used to fill in silently: edge cases, error messages, fallback behaviours. But shipping faster does not confer ownership — without access to customers, data and arbitrations, what is handed over is a burden, not a power. Product profiles move down towards building, and there too a condition appears: what counts is not the original discipline but the rung reached, and the gap between profiles turns on access to the material rather than on aptitude.

There remain the two mechanisms that keep this movement from becoming disorder. The first is structural: guarantors hold the standards of their domain, and it is their presence that authorizes the others to spill over. The second is gradual: the perimeter opens up by zones of increasing risk, the way you progressively widen a junior developer's ground, and the contribution review becomes the place where the organization says what it agrees to receive.

What the whole brings into view: the minimum level rises everywhere. The initial production of a specification, an analysis or a mockup stops differentiating anyone, and what stays scarce sorts into three families — judging what should be built, building all the way to a usable artifact, guaranteeing that a domain holds up. The responsibilities the tooling makes central — evaluating probabilistic outputs, maintaining a product memory, keeping a repository readable — belong to the third, and nothing requires turning them into posts.

Tensions / contradictions

The main tension bears on the title. This angle holds that what matters is responsibility and not the job title; a position already written elsewhere argues the reverse, namely that the title distributes the mandate ahead of any negotiation and therefore decides what its holder will be able to settle. The two meet on one point — a title without a mandate produces nothing — and diverge on the order: should you name the responsibility first, or the title that will make it credible?

Second tension, on the split between profiles. The current advantage of technical profiles rests on the state of reliability of the tools, and it could reverse: if agents become markedly more reliable, the ability to articulate a human or business reality in detail would regain the advantage over the ability to read what is going on under the hood. Nothing allows us to date that reversal, or to say whether it will happen.

Third tension, between two requirements of this angle. Multiplying contributors on the same product increases the load on the guarantors, at the precise moment when they are expected to stop being checkpoints. The setup does not say how a guarantor absorbs that load without becoming a bottleneck again.

Questions

  • At what level of risk does an organization stop opening up a perimeter, and is that threshold decided product by product or area by area?
  • How does a team establish that a guarantor has set a standard explicit enough for contributions to conform to it without consulting him?
  • What becomes of the profiles that move up neither towards judgment nor towards building, in a profession whose minimum level is rising?