Pinned Tweet
TuckerFox was already a working food delivery product. I had four weeks, I was the only designer and I worked directly with the founder.
I wrote the full case study here, including what changed, what shipped and what I would still change today.
8
4
1,482
Get more return from the features you already paid to build, with one simple UX change.
Users are less likely to try a new feature when the cost of trying it feels too high. If you keep rewatching the same show on Netflix (instead of trying a new one) you might recognize this:
A new show costs you around 20 minutes before you know if you like it. A new song on Spotify, however, only costs a few seconds.
So make trying something new feel like a small commitment. For example, if you build a new savings feature, let customers see how much they could save before asking them to set up a plan.
Make the first step easier, and more customers get a chance to discover the value you've already built. When that value brings them back, your development investment starts paying off.
9
I gave Claude a task while working on a conversational shopping AI experience for a client.
It made up a bunch of products to test the shopping flow. Now I’m more impressed with the fake products than anything else.
21
I took one client website from brief to working pages in AI.
By the end of this article you won't be vibe designing or vibe coding anymore, you'll be designing with AI the same way you design in Figma or Photoshop.
You may have read different takes on what happens to us when anyone can build software. This time I tried to look at it from another angle, keeping realism in and pessimism out of it.
407
Read the success stories you see here (including the "you can do this with AI" ones) knowing that 90% of them are exaggerated for engagement and clicks or simply made up.
A lot of us get the dopamine hit from reading them and stop there. Some want to try but have no idea where to begin.
The best way is still to learn by doing. Find the smallest useful step you can take, take it and build from there.
That's how you find out what actually works.
1
1
48
Codex burns unnecessary tokens when I do heavy UI design and development work.
To mitigate this, I started using a simple prompt-based approach. It significantly reduces token usage.
I shared the prompts in the article.
1
149
i've had enough of auto-generated ai comments that take a small, irrelevant part of my writing and make a silly comment about it.
16
I'm sure you'll see tweets like this today:
“Jev founder Diogo Almeida (@CompleteSkeptic) just put out a PDF on building a Jev harness for coding agents.”
A couple of corrections before this gets repeated everywhere.
First, Diogo didn't write the PDF.
It's an independently compiled synthesis based on design notes by Diogo Almeida. The document itself explicitly says it's not affiliated with or endorsed by TypeSafe.
The ideas are still very much worth reading. The repo also includes the source notes the synthesis was based on.
Full compiled PDF: github.com/yibie/jev-enginee…
60
I open-sourced gmail-jev, a Jev-powered Gmail tool I built for myself.
No coding needed. If you can use ChatGPT, Claude or similar, that's enough. The setup is in plain English.
It uses Jev to make quick, structured decisions about each email and returns the result as data instead of a long AI response. The app then labels your inbox and moves out what doesn't need your attention. It doesn't delete anything.
Attachments are not sent to Jev, and the decisions are saved locally as JSONL.
Less sorting. Shorter inbox. Runs on your computer.
1
1
136
See gmail-jev on GitHub:
github.com/forestwas/gmail-j…
48
JEV-GMAIL LABELER WITH CONTEXT: The labels have been color-coded, making them much more suitable for eye scanning. Also the label system is improved.
Gmail gives every email the same weight. A client waiting on a decision and the third Yandex calendar reminder of the morning sit in the same kind of row and you do the sorting in your head every time you open the inbox.
My second Jev project takes that job over, with a script running locally on my Mac.
Sensitive details are removed locally first, then Jev scores the cleaned thread for category, reply/action need, urgency and revenue relevance and Python turns those scores into Gmail actions.
Jev grouped more than 2,000 old emails in a few minutes and fine-tuning the results took only a few more. Now it checks new emails every 5 minutes and sends what the system needs as JSON.
30
Gmail gives every email the same weight. A client waiting on a decision and the third Yandex calendar reminder of the morning sit in the same kind of row and you do the sorting in your head every time you open the inbox.
My second Jev project takes that job over, with a script running locally on my Mac.
Sensitive details are removed locally first, then Jev scores the cleaned thread for category, reply/action need, urgency and revenue relevance and Python turns those scores into Gmail actions.
Jev grouped more than 2,000 old emails in a few minutes and fine-tuning the results took only a few more. Now it checks new emails every 5 minutes and sends what the system needs as JSON.
I just built my first Jev-powered task prioritization system.
My app takes today's open tasks, filters out blocked/deferred work, scores deadlines and explicit priorities in code, then uses Jev for the fuzzy parts:
- Is this actionable today?
- How much does it move a commitment forward?
- Does it unblock anything? Is it worth a focus block?
Then it picks the 3 things I should focus on today.
2
2
120
I audited a product. Then I audited the audit of the same product.
Here's what held up:
1. Severity is not priority.
A serious issue affecting a rare edge case may still matter less than a moderate issue sitting in a core flow.
2. A heuristic evaluation is not a checklist.
You don't need to force a finding under every principle just to cover all ten.
3. Best practice is not evidence.
"Most products do it this way" is not enough. You still need to explain what becomes harder, slower or more confusing for the user.
4. The same problem can violate multiple heuristics.
That doesn't make it multiple problems.
5. Desktop and mobile need separate evaluation.
The same flow can work well on desktop and create completely different usability issues on mobile.
6. UX writing is part of usability.
"This field is required" is not useful when the actual problem is an invalid email address.
7. A finding needs observable user impact.
Naming a law, principle or pattern is not enough. The report should explain what actually goes wrong for the user.
8. Every finding should be defensible.
Before reporting an issue, I want to be able to reproduce it, show the evidence, explain its impact and clearly separate the observation from the recommendation.
Heuristics are useful because they give you a lens. What you do with that lens is judgment.
34
















