Building What WCAG Can't Test

I'm neurodivergent. Software usually isn't designed for people like me, and for a long time, I couldn't build my own because you had to already know how to code to get in the door. Dyscalculia meant that was a huge, ‘Nope,’ more often than not, because most math and I just don’t mix. Oil and water, and then some.

So in 2021, I did the next best thing: I showed up to the WCAG cognitive accessibility working group meetings and contributed feedback on the standards. If I couldn't build the software, maybe I could help shape what good was supposed to look like.

Then I burned out. Day job, life, the usual. I stopped going. I actually ended up taking three weeks off at the end of 2021 due to a mental health crisis. It’s not an exaggeration to say I almost wasn’t here to write this. Thankfully, I made it through thanks to my amazing, supportive co-parent and a weird little cartoon. Also, if you’re taking depression medication and ADHD meds, worth checking how those interact. Ask me how I know!

This year, I started building again using the Zero Vector Investiture framework as a jumping off point, and I finally am building the software I always wanted to exist. And now I'm looking at what I built and comparing it to the guidelines I helped shape. Here's what I’ve found so far:

The WCAG cognitive accessibility guidelines, more specifically the "Making Content Usable" document, are the best framework we have for designing software that works for people with cognitive and learning disabilities. They define 8 objectives and 58 design patterns: things like "use plain language," "avoid interruptions," "don't rely on memory," "provide feedback on progress." This is all good, important stuff to be talking about. A lot of software fails on basics that COGA has had answers to for years.

But COGA was designed for a specific model of accessibility: Stable impairments, stable environments, discrete tasks. Fill out a form. Navigate a website. Understand a document. The user has a cognitive disability, and the goal is to remove barriers to completing the thing.

That's not wrong. It's just not the whole picture.

The first thing COGA doesn't account for is that cognitive capacity isn't fixed. It fluctuates by time of day, by what just happened in Slack, or at your office, by how many social interactions you've had, by whether you slept. The guidelines assume a stable accessibility barrier in a stable environment. But for a lot of neurodivergent people, the profile changes hour to hour.

So I built this into Lumentide, a task companion I’m developing focused on neurodivergent minds. The app assesses your current energy using a combination of time-of-day heuristics (morning defaults to high, afternoon and evening to low) and manual override (because time-of-day won't be right for everyone), and the override is always one tap away. It also tracks social battery: If you've logged a two-hour meeting and your capacity has dropped below a threshold, the energy assessment shifts down automatically, regardless of what time it is.

Every task in the app carries an energy tag: Low, medium, or high, either set manually or guessed from the task's language when you capture it quickly ("reply to" reads as low; "build" reads as high). When you open the app, you see two or three tasks that match your current energy level, not your full list. Low tide shows you low-energy tasks only. High tide opens everything up. The ones you've already started float to the top, because momentum is its own kind of energy.

Image description: Three task suggestion groups organized by energy level. Low Tide: Reply to that email, Water the plants, Stretch for 5 minutes. Mid Tide: Draft project outline, Organize desk, Make that phone call. High Tide: Deep work session, Tackle hard problem, Write important document. Each group is color-coded with a left border: Blue, tan, and amber.

That's a different design problem than COGA is solving. COGA helps you remove barriers to completing a task. Lumentide helps you figure out which task you can actually start.

The second gap is about completion. COGA's error-recovery patterns assume finishing a task is the goal and obstacles are the problem. For a lot of ADHD brains that's backwards: The obstacle is starting, not finishing. Getting from "I should do this" to "I am doing this" is the hard part.

In Lumentide, this is handled by a character called Citrine the Start Celebrator. When you tap the start button on a task, Citrine appears with a single line: "You started. That's the hardest part." No points, no streak counter, no gamified response (and no guilt tripping birds, sorry other apps. What is it with all the various guilt tripping birds??) Just acknowledgment that initiation is the actual achievement. The task then gets marked "warm", a status that floats it to the top of your suggestions for the rest of the day, because momentum deserves to be sustained, not just noted.

Image description:  Lower spirits grid showing Vesper, Arcana, Citrine, and Amethyst with their full names revealed. Vesper tracks social energy and twilight approach. Arcana prevents task-switching whiplash. Citrine celebrates starting, not finishing "The spark is the hardest part." Amethyst notices patterns without judgment. Aquaria the Rest Keeper watches battery levels and suggests recharge: "Rest is data, not weakness."

The warmth system matters for a reason beyond encouragement. A lot of ADHD brains will start something, get pulled away, and lose the thread entirely. A warm task is a signal back to yourself: this is where you were, you don't have to start from scratch. COGA has error recovery, undo, and progress indicators are all designed for a user completing a linear task. It has nothing for a user trying to re-enter something they already started after their attention went somewhere else. That's not a completion problem. It's a re-initiation problem. It's not in the spec.

The third gap is task-switching. COGA addresses cognitive load within a task: Simplify the interface, reduce distractions, chunk the content. What it doesn't address is the cost of switching between tasks. For ADHD and autistic users, finishing one thing and starting another isn't neutral. It's a dysregulation event. The brain needs a buffer.

In Lumentide this is Arcana, the Threshold Guardian. When you complete a task, Arcana surfaces a short transition ritual that lasts two to five minutes (it’s your choice on the task transition duration) before the app presents the next thing. A brief walk. Some water. A breathing exercise. Something that creates a seam between what just ended and what's about to begin. The transition also logs as a recharge event, which feeds back into the social battery model: finishing something and moving on isn't treated as neutral, because it isn't.

COGA's guidance at task completion is "provide feedback on task success", which is usually a confirmation message, a progress update. For a lot of neurodivergent people, the moment after finishing something is a vulnerable transition, not a tidy endpoint. The spec has no pattern for what happens in between. Arcana is designed entirely for that gap.

The fourth gap is grief. I want to be direct about this one: Grief is not a cognitive disability. But acute grief disrupts cognition in ways that parallel cognitive impairment: Difficulty concentrating, memory fragmentation, sensory sensitivity, impaired decision-making. Temporary, yes. But real, and poorly served by software. This is where I’m at right now.

The specific kind of grief I'm researching is ambiguous loss: Mourning people who are still alive but absent. A parent with dementia. An estrangement. A relationship that ended but left someone still present. Dr. Pauline Boss, who named the concept, argues this kind of loss is particularly disorienting because it has no ritual container. There's no funeral, no casseroles, nobody calling to check up on you because they don’t know what to do when you’re mourning someone who’s still alive. Most people don’t know what to do when your grief has no socially legible endpoint. Even worse, your brain keeps getting interrupted by a grief it has no script for, with no real ‘end’ in sight, and there might not ever be one.

Image description:  Foghorn ritual capture panel. A 1–10 intensity scale asks "How present is the grief right now?" Below it, loss type buttons: A person, A relationship,      Yourself, A place, All of it. A text field asks "What do you remember?" Record Ritual and Not now buttons at the bottom. A historical echo reads: Echoes of Falmouth, February 1985, Clouds, 44°F, W wind.

That’s where Foghorn comes in. Foghorn is a tool that uses daily and historic weather patterns as a ritual anchor for grief processing when accompanied by certain sounds. Each morning: You can check the weather, and play an optional foghorn sound if the conditions match (fog, rain, snow, the kinds of weather that already carry emotional weight, at least for me), record how present the grief feels on a scale of 1 to 10, note what's with you. This is pretty heavily personalized to me right now, for the grief I’m processing and sitting with, but I have to believe it still has value, or will to others once it’s live…because I know I’m not the only person processing this sort of grief and loss. Over time the app builds a record: Intensity trends, what loss type is surfacing, how grief oscillates across phases. The foghorn sound is a real recording from a lighthouse in my hometown. The specificity matters: It's a real place, a real sound, not a UI element or something I pulled off of Storyblocks and threw into the app so it ‘made a cool noise’ when I clicked a button. I wrote a whole thing about it on LinkedIn, it’s messy to the point of oversharing, but hey. That’s life.

Image description:  Foghorn's main screen over a grey coastal background. A weather card shows: Clouds, 48°F, feels like 41° in Cut-off Junction. Below it: a quarterly reflection notice, a grief phase selector set to Not tracking, a Play Foghorn button, a Record Ritual button, and a historical echo: Falmouth, February 1985, same cloudy conditions.

COGA has no guidance for any of this. Not for emotional content that requires pacing rather than efficiency. Not for ritual as an accessibility pattern, the idea that some interactions should be slow, repetitive, and sensory on purpose. Not for software whose goal is to help someone move through a state rather than complete a task. The framework wasn't built for it, and I don't think that's a flaw, it's a scope boundary. But grief is a design context that millions of people are in at any given time, and right now there's no WCAG guidance for what it means to build software for them well.

Image description:   Foghorn app Field Notes panel floating over a rocky coastal shoreline in muted grey tones. The empty state reads: "Field notes are longer reflections: Researcher memos, observations, anything that doesn't fit in a ritual note." A plus New note button sits in the header. The footer credits the Nobska Lighthouse foghorn recording.

The fifth gap is social energy. In Lumentide, this is Vesper, the social battery watcher. You log your social interactions as they happen: A meeting, a phone call, a difficult conversation. Each carries a drain rate. The battery percentage is visible at a glance, and it feeds directly into the energy assessment: When the battery drops below 25%, the app shifts to low-energy task suggestions automatically, regardless of what time it is. The recharge side is tracked too: A solo walk, quiet work, time outside all log as recharge events, and a separate spirit surfaces suggestions when the battery is critically low.

COGA as it exists today has no concept of the interaction as a cognitive cost. Its user model is a person at a screen trying to complete a task: The social context surrounding that moment is outside the frame. For a lot of neurodivergent people, a two-hour meeting isn't just a meeting. It's an accessibility event that degrades everything that comes after, for hours. Software that doesn't model that is optimizing for a user who doesn't exist.

The sixth gap is the hardest to name. In Lumentide, every task carries context from the moment it was created: the time, the energy level, the social battery percentage. Over time that becomes a behavioral record, when you create high-energy tasks versus when you actually complete them, how social load correlates with output, what your real working patterns look like versus what you planned. The app surfaces this back to you. The design intent is therapeutic self-awareness: not optimization, not productivity metrics, but helping someone understand their own rhythms well enough to work with them instead of against them.

This is where the real gap opens. Behavioral summaries are a reasonable  accommodation. What I built goes further: It's a behavioral record, and the people it's built for have historically been surveilled, diagnosed, and pathologized based on exactly this kind of data, collected by systems that didn't ask (case in point, my 3” tall stack of medical records and testing I didn’t consent to as a child, that got me the diagnoses I have today. Grateful, yes, but the impact and the ‘therapy’ surrounding it was traumatic). So I made different choices: Local data storage first, no external transmission, the data never leaves your device. Not as a technical footnote, but as a design principle. If you're building tools for neurodivergent users, you don't get to be casual about what you're collecting, or at least you shouldn’t be.

None of this is a criticism of WCAG or COGA, certainly not. The people working on it are doing important, largely unpaid work within a standards process that moves slowly by design. The guidelines exist because some amazing people fought to put them there.

But standards follow practice. The patterns that end up in WCAG start as things people noticed in the real world, discussed, documented, and brought to the table. That's how the authentication patterns in WCAG 2.2 happened. That's how the cognitive accessibility work I contributed to in 2021 happened.

The gaps I've described aren't hypothetical. They're documented in working code: fluctuating capacity, protective incompletion, transition regulation, social energy as a cognitive cost, grief as a design context. That's the kind of practice that belongs at the standards table. Not as a critique of what COGA got wrong, but as evidence of where the frontier is, and where I’m going as I continue to grow in my career, my goals for my life, and the lives of those around me. The future of design works with human rhythms, not against them, and I'm building it. I hope you’ll join me along the way.

Image description: System diagram showing Human-Rhythm Design: A framework spanning Cognitive, Temporal, and Emotional dimensions. Four branches: Memory (Everbloom, reading and context support), Attention (cognitive UX pattern library and emotion/grief rituals), Time (Lumentide circadian UI and interruptible workflows with autosave), and Transitions (Foghorn grief rituals and support)