The Great Role Confusion of 2026

Happy Friday!

Bell Labs introduced the "Member of Technical Staff" title in the mid-20th century (well before Xerox PARC was founded in 1970). It was held by legendary figures in computing and physics, including Unix co-creator Dennis Ritchie when he joined in 1967.

Thanks for reading Peter's Context Design! Subscribe for free to receive new posts and support my work.

Bell Labs deliberately designed the title to serve specific cultural goals:

• Flattening Hierarchies: Instead of using rigid corporate rankings like "Junior Engineer" or "Principal Engineer," the broad MTS designation was applied to researchers and engineers alike.
• Promoting Fluidity: It allowed employees to move seamlessly between writing code, constructing experimental hardware, and doing pure mathematical or scientific research without being pigeonholed by a narrow job description.
• Academic Fellowship Vibe: It functioned more like a scientific research fellowship than a corporate ladder step, emphasizing collective expertise.

AI and research-heavy tech organizations like OpenAI, Anthropic, and Google DeepMind have adopted that title.

So should UX adopt something similar? Member of the User Experience Staff? It doesn't seem that crazy right now.

MC Dean at Google wrote about A Constellation of Emerging Design Roles, looking like this:

1. Design Engineer
2. AI Experience Designer
3. Director of AI Design and Experience
4. Computational Designer
5. Prompt Designer
6. AI Designer
7. Orchestrator (or AI Orchestrator)
8. AI Systems Designer

Boris Cherny at Anthropic came up with these:

1. Prototyper: comes up with brand new ideas; churns out many ideas, most of which don't ship
2. Builder: quickly turns a prototype/idea into production-grade product/infra
3. Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance
4. Grower: takes a product that has been built and iterates on it to improve Product-Market Fit
5. Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales

I've written a bit about this on my blog, calling it the Great Role Confusion of 2026.

My thinking on these has been:

1. Context Designer. Context engineering is about creating systems that pull in all the right context. Context design is about understanding humans and usage and figuring out what the context is that needs to go in there.
2. Forward Deployed Experience Designer. Large amounts of roles and processes are changing with AI. The way the shape of that change is discovered right now is with Forward Deployed Engineers, because they can prototype and implement change. I'll argue that UX design has a TON of experience (see what I did there?) to add to this.

Lots to be figured out! I'll say one thing though: there is power in defining new roles. More on that later.

👀 Meanwhile:

A very long post on The Layers of AI UX. Not a bad stab at layering out this new world, as we UX people are wont to do. My attempt has been model, context and experience.

MC Dean at Google on how design roles are fanning out: A Constellation of Emerging Design Roles:

These include Design Engineer, AI Experience Designer, Director of AI Design and Experience, Computational Designer, Prompt Designer… (more here). One writer at Fast Company called them "Frankenjobs," roles stitched together out of three old jobs and a hope.

No Figma. No Jira. No docs. Gusto CTO Eddie Kim: 4 engineers and 1 designer built Gusto Cofounder, zero to a tier-one launch in 10 weeks, no PMs. Their "trash can method": write the full PR, review it, delete it. That's the product decision, no planning doc. A designer with no engineering background hit the 94th percentile for shipping code.

Taras Bakusevych compiled 39 principles for designing human-AI interaction, pulling from Horvitz, Microsoft's HAI guidelines, Google PAIR, and Anthropic's Constitution.

Victor Yocco's Task Audit framework argues against defaulting every AI feature to a chatbot. Case study: utility technicians on high-voltage lines, gloved hands, bucket trucks, got voice input and audio output instead of a touchscreen. Diagnostic time dropped 20%.

Heenesh Patel on why systems thinking is becoming the core UX skill:

"In the platform driven experience, the systems thinking designer has the clear advantage."

Ethan Mollick on the twilight of the chatbots: a quarter of OpenAI staff run four or more agents weekly, and legal and HR adopted agents at nearly the same rate as engineers. A separate study of Claude Code users found domain expertise, not job title, predicted how useful people found each prompt.

Gale Robins: discovery is a capability, not a phase. AI speeds up the execution side of product discovery, not the judgment side. Draws on Argyris's double-loop learning: document the reasoning behind a decision, then check it against what happened, instead of moving straight to the next release.

Someone ran a public challenge to see if anyone could leak secrets from an OpenClaw assistant by emailing it. After 6,000 attempts and $500 in token spend, nobody got through.

Saeideh Bakhshi on the maintenance layer: as first drafts get cheap (a plan, a name, a rough concept in an afternoon), tending and revising after the novelty wears off becomes the differentiator.

"The first version was never the whole work."

Charity Majors: the case for AI mandates. This is really good.

Adrian Levy: you don't design the interface anymore, you design the deciding.

Slava Polonski argues AI personality is a design problem, currently an accidental byproduct of alignment training that should be a deliberate design choice. I have to disagree. AI personality is carefully designed in post training.

NN/g on establishing baselines for impact: teams that skip a baseline before shipping can't later prove the work mattered, even when it did.

The Strategic Linguist dissects the LinkedIn "executive departure" post as a genre anyone can borrow regardless of seniority: gratitude, a survey of "the macro," a declared calling, credit to unnamed teams. Cites a new Cornell study, the Corporate Bullshit Receptivity Scale: people most impressed by AI-generated corporate jargon scored worse on analytical thinking and cognitive reflection than people who saw through it.

Patrick Neeman applies Brandolini's Law to generative AI: refuting bullshit always cost ten times more than producing it.

"The amount of energy needed to refute bullshit is an order of magnitude bigger than to produce it."

Michael Buckley on when the profession outruns the mentor: mentorship assumes a profession changes slower than a career unfolds.

From my own site: Attention Residue, quoting Sophie Leroy on why part of your attention stays with the last task instead of fully moving to the next one.

"As we switch between tasks, part of our attention often stays with the prior task instead of fully transferring to the next one."

Also on my site: Skill Erosion.

Claire Vo's Lenny's Newsletter podcast: a Sonnet 5 review, 64 generations in.