ebunalabi — field notes
~ field notes
...on accessibility, developer tools, low-code builders, and the unglamorous craft of production-grade compliance + the personal stuff: career, life, the long arc.
timeline
Honestly, I think Figma will be here for much longer than the market thinks. Agentic workflows are simply not enterprise ready, too non-deterministic, skill level varies greatly across board and so outcome vary a lot. Enterprise is probably better off sticking to its already set workflows, versus having 10,000 engineers plus designers attempt to produce something useful using agents that are incredibly difficult to guard.
Currently navigating a form of burn out. Not sure how to move from here yet, but typically I’d take a break and do something I enjoy. This year feels like it’s been 12months already but yeah I’d take it easy.
Presented the research findings of a usability study I’ve been working on for the last 4 months! But this is just phase 0, next will be to get these new requests on the right backlogs and get those committed and developed. The beauty of having customer research before development is that now, we are going to build with a clear understanding of customer pain points. Validation still has to happen after the fact, but for phase 1 we have lots of evidence.
Couldn’t agree more with Cat Wu’s point on Lenny’s pod, build things you use everyday, it’s really the best way to develop “product taste” which is probably going to be the key distinguishing characteristic of engineers and PMs that survived the AI-pocalypse.
This is another article we referenced in the chat about FIRE, life is short and creating memories outside of work and mundane stuff is critical.
Life is Short by Paul Graham
paulgraham.comvisit
In a previous voice note on this timeline we discuss why everyone should have a hobby, Obinna referenced this video so I'm sharing here.
Hobbies: Your Best Defense Against Brain Rot
Dr. Izzy Sealeywatch
I fear companies might start rolling back token access across different orgs, the usage is definitely bloated by a lot of inessential tasks likely more efficiently and productively completed without an AI agent.
All AI agent operators need a visual editing layer that allows users edit placement, text, alignment etc. I’ll rather move the elements around myself rather than ask an agent to do that, efficiency gains are sometimes lost in the back and forth. This ties back to the need for deterministic workflows.
Something I didn’t realize was critical for my productivity was actively blocking out time to think. So this year my calendar has had fixed focus time, where I get real work done. No meetings, no nothing, blocked months in advance. Product Management has a way of keeping you busy attending meetings yet no meaningful input or contribution is being made.
One of the biggest skills I’ve had to learn as I navigate my career as a Product Manager working in a pretty huge org is; always define ownership of projects/tasks. Also getting comfortable with asking my manager for clear expectations. Pushing back on projects that I think isn’t high value or doesn’t have leadership buy-in was something I didn’t get started on early enough.
What does AI native operational efficiency really entail for engineering orgs in terms of tokenization and relative ROI?
Traditional companies ( companies with definite and specific roles that contribute to different parts of the SDLC ) still need deterministic workflows to support the scale at which features are being shipped across multiple product verticals. The lethargy to shift entire teams and orgs to AI first and only workflows is fair and understandable.
The AI native migration friction that traditional companies are experiencing is real. AI ready companies are able to hire versatile technical staff, traditional companies are really struggling to manage the synergy across roles especially now that AI usage has morphed multiple roles into one. Efficiency and ROI is still a big question !
On the way back from a call. Compliance is the only PM job where the user you're protecting is usually not in the room.
“Guardrails are love.”
- note7 min read
Compliance is not a checkbox
Compliance failures don't come from missing controls. They come from controls that exist on paper but don't exist in the workflow.
“Most accessibility wins are friction wins for one specific persona that quietly turn out to help everyone.”
If your low-code builder doesn't let users export their own data, it's not a builder. It's a hostage situation. Export is a feature you ship on day one or you owe your users a really good apology.
Subscribe
subscribe
Occasional digests of new notes and audio. No tracking, easy unsubscribe.