The most uncomfortable conversation in design leadership in 2026 is the one where a Staff designer asks, with real fear behind it, "if the model makes the screens now, what exactly am I senior at?" The honest answer is that the senior IC's job did not disappear; it split into three modes that the old job description never named, and the leveling rubric you inherited measures none of them. A senior IC now spends their week curating which AI outputs are worth pursuing, editing generated work up to the bar, and authoring the original moves a model cannot make - and what separates a Staff from a Senior is no longer how fast they push pixels but how much of their time is spent in authorship versus the lower modes. This lesson reframes the role around those three modes, names the time-split that distinguishes the levels, and hands you a revised IC ladder that makes the new job legible, promotable, and defensible to the very people who fear it.

Why the Old Ladder Broke

The IC ladder you inherited measures the wrong things, and it broke the moment generation got cheap. Most ladders describe seniority in terms of scope and autonomy: a Senior owns a feature, a Staff owns a surface, a Principal owns a product area. Underneath, the implicit measure was production - a senior designer was someone who could produce more, faster, and better than a junior. When production was the scarce skill, that proxy worked. It does not work now, because the model produces the first eighty percent of any artifact in seconds, and "produces good screens fast" no longer distinguishes anyone. A junior with a good prompt and a senior with thirty years of taste both get a competent first pass in nine seconds. The ladder's hidden axis went flat.

This is why senior ICs feel the ground move even when their title is safe. Their value was real, but the ladder described it through a proxy (production) that AI commoditized, leaving them unable to articulate what they are senior at. They reach for "taste" and "judgment," which are true but unmeasurable, and the conversation dies in vagueness. The fix is not to insist that senior designers are valuable; it is to name the three modes the work actually splits into now, measure seniority by the mode-mix rather than by production volume, and rebuild the ladder so a designer can see their own growth in terms the new job rewards.

Naming the three modes does something else important: it gives the anxious Staff designer a vocabulary for their own value. "I am senior because I spend most of my week in authorship and high-stakes curation, and I can move junior designers from mostly-editing into curation" is a sentence a person can say in a performance review, defend in a promotion case, and use to direct their own development. Vagueness offers none of that. The modes are the vocabulary that converts unmeasurable dread into a measurable, growable skill.

Mode One: Curation

Curation is the work of deciding which generated outputs are worth pursuing and which to discard, and at what level the decision is made. When a model produces forty onboarding concepts, curation is keeping six with reasons and killing thirty-four with reasons. When a PM brings three Lovable prototypes, curation is deciding which direction the team commits to. When the design system could go three ways, curation is choosing the path the org will live inside for years. Curation is the highest-leverage mode because a single good curation decision determines the value of everything produced downstream - a wrong direction kept means a hundred hours editing the wrong thing.

The reason curation is now a named mode rather than an invisible reflex is volume. When you produced one concept, there was nothing to curate; you made it and shipped it. When the model produces forty, the bottleneck moves entirely to selection, and the skill of selecting well - against user intent, against strategy, against what the org can sustain - becomes the work itself. Curation is also where seniority shows most clearly, because a junior curates on surface appeal ("this one looks nicest") while a senior curates on consequence ("this one looks worst but is the only direction that survives the edge cases and the brand"). Same forty concepts, completely different keep-list, and the difference is exactly the judgment the model lacks.

The Curation Trap

The failure mode in curation is anchoring - accepting the model's first or most polished output because evaluating forty options is exhausting and the first one looks fine. Senior curation resists the anchor deliberately: it treats the model's most confident output with the most suspicion, because confidence and polish are exactly what the averaging machine produces reliably regardless of correctness. The senior IC's curation move is to ask "which of these is right" rather than "which of these is finished," and the gap between those two questions is most of what curation seniority means.

Mode Two: Editing

Editing is the work of taking a generated artifact that is roughly right and bringing it up to the bar - fixing the hierarchy the model got balanced-but-wrong, correcting the destructive-action placement, adding the error and empty and loading states the model skipped, aligning the tokens that drifted, finishing the responsive behavior. Editing is where most of the hour-by-hour work of an AI-augmented senior designer now lives, and it is genuinely skilled work, because knowing what to fix and in what order is itself judgment. But editing is the middle mode for a reason: it is necessary, valuable, and yet not where a senior IC's distinctive value concentrates, because a strong mid-level designer can also edit competently.

The trap in editing is that it is endless and seductive. There is always another generated artifact to bring up to the bar, and editing feels productive because you can see the before and after. A senior IC who spends ninety percent of their week editing has effectively been demoted to a high-end finisher of machine output, no matter what their title says, because they have stopped doing the two modes that actually require seniority. The ladder has to make this visible, because the org will happily let a Staff designer edit all day - it is useful work - while the designer quietly loses the authorship and high-stakes curation that justified their level. Editing is a tax on the value of generation; it is not the value itself.

The senior move within editing is to make editing teachable and delegable - to turn "here is what I fixed and why" into a verification rubric a mid-level or junior designer can run, so that the senior's editing capacity multiplies through others rather than being consumed one artifact at a time. A senior who edits is doing the job; a senior who turns their editing into a transferable standard is doing the senior version of the job, and the ladder should reward the second far more than the first.

Mode Three: Authorship

Authorship is the work of making the original moves the model cannot make: the novel interaction pattern that has no precedent in the training data, the brand-defining visual decision that resists the model's convergence, the structural insight from research that reframes what the product should even be, the system-level decision that shapes everything generated downstream. Authorship is rare, high-stakes, and exactly the low-frequency quadrant where models are unreliable - which is precisely why it is the mode that most defines senior value in an AI-native org. The model cannot author because authorship requires a model of the user, of the brand, of the strategy, and of what does not yet exist, none of which a probability distribution over past designs contains.

Authorship is also the scarcest of the three modes and the easiest to lose track of, because it does not feel productive in the way editing does. Authorship looks like staring at a wall, redrawing the same flow eleven times, killing your own good ideas, holding a problem in your head for three days before the right structure appears. There is no satisfying before-and-after, no queue cleared. An org optimizing for visible throughput will starve authorship without meaning to, by filling the senior IC's week with editable generated output until there is no room to author anything. Protecting authorship time is a deliberate act, and the ladder has to name it as the thing senior ICs are most accountable for, or it evaporates under the gravitational pull of the endless edit queue.

The distinction that matters: authorship is not "design without AI." A senior author uses the model constantly - to explore the space around an original idea, to pressure-test it, to produce variations on a move the human originated. The model is a collaborator in authorship. What makes it authorship rather than curation or editing is that the originating move - the thing the variations vary around - came from the human, from a place the training data does not reach. The senior IC authors the seed; the model helps grow it; the senior curates and edits the growth. All three modes braid together in a single piece of work, which is why the time-split is about emphasis, not separation.

The model commoditized production, so production stopped distinguishing seniority. What distinguishes a senior IC now is the mode-mix: how much of the week is spent authoring the moves a model cannot make and curating at consequence, versus editing machine output up to the bar.

The Time-Split That Distinguishes the Levels

Here is the reframe that makes the ladder work: levels are distinguished not by scope alone but by the mode-mix, the proportion of time spent in each of the three modes. The numbers below are illustrative anchors, not a stopwatch - the point is the shape of the shift, which is that as you climb, editing falls and authorship plus high-stakes curation rise.

A mid-level designer in an AI-augmented practice spends most of their time editing - bringing generated output up to the bar - with some curation (mostly at the artifact level) and occasional authorship under guidance. Their week is roughly editing-heavy, and that is appropriate; editing well is how they build the judgment that authorship requires. A Senior shifts the balance: still substantial editing, but now leading curation on their surface and authoring the key moves within it, with their editing increasingly turned into rubrics others run. A Staff tips past the midpoint - authorship and consequential curation now dominate the week, editing has largely been delegated as teachable standards, and the designer is accountable for the original moves and the direction-setting selections across a whole area. A Principal is mostly authorship and the highest-stakes curation - the system-level and strategy-level decisions that shape what the entire function generates - with editing essentially gone from their personal week and present only as the standards they author for everyone else.

The shape is the lesson: seniority in an AI-native IC role is the rising share of authorship and consequential curation, and the falling share of editing. This is measurable in a way "has good taste" is not. You can look at a senior IC's actual week and ask what fraction was spent authoring versus editing, and if a Staff designer's week is ninety percent editing, that is a development conversation, not a mystery. The mode-mix turns the invisible new job into something a manager can coach toward and a designer can grow into.

The Revised IC Ladder

The artifact is a revised IC ladder, and it differs from the one you inherited in three structural ways. First, each level is defined by mode-mix, not production volume. The level descriptions name the expected balance of curation, editing, and authorship, and explicitly state that "produces high-quality screens quickly" is table stakes at every level rather than a differentiator at any. Second, each level names the curation stakes it owns. A Senior curates at the surface, a Staff at the area, a Principal at the system and strategy - the consequence of the selection decisions rises with level, which is how curation becomes a promotable axis rather than an invisible reflex. Third, the ladder names authorship as the protected scarce resource at senior levels and makes "protected authorship time and original moves shipped" an explicit expectation, so the org cannot quietly demote a Staff designer into a finisher without it showing up as a ladder failure.

The ladder also adds a column the old one lacked: how editing scales. At junior levels, editing is done; at senior levels, editing is turned into transferable rubrics and standards that multiply through the team. This makes the difference between a Staff designer who edits forty artifacts and one who authors the verification standard that lets four mid-levels edit forty artifacts each - and the ladder rewards the second, because that is leverage and the first is just labor. The revised ladder, in short, makes the three modes visible, makes the mode-mix the axis of advancement, and protects authorship from the gravity of the edit queue.

A Worked Example: Recalibrating a Staff Designer

Make it concrete with the anxious Staff designer from the opening. They came to you afraid that the model made them redundant, and a week's audit of their actual work shows why: their week is roughly eighty percent editing - cleaning up generated mocks for three squads, fixing tokens, adding states - ten percent curation, and almost no authorship. By the inherited ladder they look fine; they are productive and the squads are happy. By the revised ladder they are mis-leveled: they are doing Senior-grade editing labor at a Staff title, and the authorship and consequential curation that justify Staff have been crowded out entirely.

The recalibration conversation, with the modes as vocabulary, becomes constructive rather than threatening. You show them the mode-mix and name the problem precisely: not that they are bad at their job, but that the org has let the edit queue consume the time the role exists to spend on authorship. The development plan writes itself: turn their editing expertise into a verification rubric the three squads' mid-levels run, freeing their week; reassign the consequential curation decisions on those surfaces to them explicitly; and protect two days a week for the authorship the surfaces actually need - the novel patterns, the brand moves, the structural research insights that no model and no edit queue will ever produce. Six months later, their week has inverted toward authorship and high-stakes curation, their value is legible to them and to the org, and the redundancy fear is gone - not because you reassured them, but because you gave them a job whose seniority a machine cannot commoditize and a vocabulary to name it.

Putting It to Work This Quarter

Audit the mode-mix before you touch the ladder. Pick three senior ICs and look at their actual week - what fraction went to curation, editing, and authorship - and you will almost certainly find authorship starved and editing bloated, because that is the default an AI-native org drifts toward. The audit is the evidence that the role has quietly changed under everyone, and it is what makes the ladder revision feel like a fix rather than a fad.

Then write the revised ladder around mode-mix, name authorship as the protected resource, make curation a promotable axis with rising stakes, and reward editing-turned-into-standards over editing-as-labor. Take it to the anxious senior ICs first, because they are the ones who most need the vocabulary, and a ladder that gives a frightened Staff designer a defensible answer to "what am I senior at" is worth more than any reorg. The deliverable is not a tidy rubric; it is a role definition that survives the commoditization of production by relocating seniority to the modes the model cannot touch - and a team of senior ICs who can finally say, with evidence, what they are senior at.

Key Takeaways

• The inherited IC ladder broke because it measured seniority through production volume, and the model commoditized production. Senior ICs feel the ground move even when their title is safe, because their value is real but the ladder described it through a proxy AI made worthless.
• The senior IC's job split into three modes: curation (deciding which generated outputs to pursue, at rising stakes), editing (bringing generated work up to the bar), and authorship (the original moves a model cannot make). The modes braid together in a single piece of work; the question is the emphasis.
• Curation is highest-leverage and shows seniority clearly: a junior curates on appeal ("which looks finished"), a senior on consequence ("which is right"). The trap is anchoring on the model's most polished, most confident output.
• Editing is where most hour-by-hour work now lives, but it is the middle mode and a trap: a senior who edits ninety percent of the week has been demoted to a finisher of machine output regardless of title. The senior move is turning editing into transferable rubrics that multiply through others.
• Authorship is the scarcest, most senior-defining mode - the low-frequency moves models cannot make - and the easiest to lose, because it does not look productive and the endless edit queue crowds it out. Protecting authorship time is a deliberate act the ladder must enforce.
• Seniority equals the rising share of authorship and consequential curation and the falling share of editing - a measurable mode-mix, not unmeasurable "taste." The revised ladder defines each level by mode-mix, makes curation a promotable axis with rising stakes, protects authorship as scarce, and rewards editing-turned-into-standards over editing-as-labor.