robertnowell

the main concern here is what is needed to create and sustain that growth rate.

here are three potential issues:

1. there is a short term incentive for lying -- tricking people can get you a long way (e.g. delve)

2. there's a genuine long term incentive for selling products that have short term benefit but long term harm (e.g. gambling, cigarettes, etc).

3. there is a durable incentive to sell products that genuinely benefit for your customers but cause net harm to society -- this last one is a hard problem of capitalism, and imo it's the gov'ts job to make sure that such companies are not allowed to win.

if they are selling to a business, the biz will pay if it solves their problem. if the solution doesn't solve their problem, or something else solves their problem that is easier / cheaper / better, the business will not pay.

their number one priority is solving their problem so they can realize their organization's mission, and that's how it should be.

if the translation is good enough to solve their problem, then it doesn't need to be any better.

the version of this skillset that stays employed is "now I translate 10000x more than i could before by managing a fleet of agents. by encoding my experienced taste and judgement into robust evals, I've helped my ai translators be far better than chatgpt on its own, and much more cost effective compared to manual human translation"

head in the sand

if it was valuable, people would pay for it

unfortunately this person will soon be unemployed.

not because their skills are no longer relevant, but because they are taking a principled stance defending now irrelevant skills.

unfortunately this person will soon be unemployed.

this is like telling a student driver they have to push their car by hand instead of putting their foot on the pedal.

fair point, but it is not a tesla-style “giant iPad”

primary inputs while driving are buttons and knobs

you are correct, I amend my claim.

the interiors are nice, but overall imo the car is a design failure.

imo the exterior looks like a cartoon BYD and should be killed with fire.

however while folks are negative on Ive for the car shape, he only designed the interiors not the car body, and the interiors are kinda lovely car interior design:

-no touchscreen (dangerous while driving) -clicky, intuitive tactile switches and buttons -thoughtful use of color (display base color changes based on driving mode)

I mean even just looking at those air conditioning vents (rotate vent to open/close) is classive Ive: intuitive but sophisticated.

I hope more manufacturers copy these new/old patterns on the interiors.

> Every software developer knows that you can’t make projects go faster just by typing faster. If that were the case we would all be taking typing lessons.

^ this statement is false. typing infinitely fast would make software development much faster.

typing infinitely fast would not make shipping useful products and features instantaneous, because there is product, technical, and organizational uncertainty that requires iteration and "cross functional collaboration" to figure out.

but ai can make each iteration step a lot faster.

interesting visualization of 51 real engineering tasks completed by openclaw ai engineers during March 2026

I was having a hard time explaining how my OpenClaw ai engineering orchestrator runs a structured engineering process by dispatching tasks to OpenAI Codex or Claude Code

...so i rendered it into a video game.

watch a team of Claude-powered ai employees compete against team OpenAI on real Kopi AI engineering tasks.

you are seeing a visualization of 51 actual tasks executed on by a team of openclaw AI engineers during March 2026

these engineers all follow the same structured process.

a task is created by me from slack, and the ai starts a structured process to execute on the task

each desk represents a sub-task dispatched to this employee's coding agent -- either claude code or openai codex

in this way we have two competing teams of ai engineers -- openai vs claude -- you can see each team's score on the left

every employee plans, builds, audits, fixes, tests, and if thye are succesful, creates a PR.

Sometimes they have to audit and fix in a loop several times before they get approval.

if their PR is successfully merged, they get a quick moment to enjoy the plants on the top floor, and then it's back to work.

Sophie is one of our star employees, but not everyone is doing so well.

Sometimes tasks don't get merged, and they fail.

this openclaw engineering system is called b0b, and it's open source on github

just made an ai graphic designer for marketing emails

learns about your brand and creates custom email graphics for headers etc

pretty cool what gpt-image-1 can do

if curious, can check out https://graphic-design.email

what’s an example of a code search problem that vscode’s native search fails to solve?

illustrated with midjourney: https://open.substack.com/pub/robertnowell/p/ai-a-means-to-o...

Vision pro + AI allows me to create a decent looking, functional user interface just by talking about it. While feeling like I'm on a mountaintop, immersed in nature.

midjourney magazine is "soulless and embodying the threat AI poses to journalism", merely "pages upon pages of large images of varying quality in varying genres... captioned with the prompt used to generate them".

The author doesn't seem to be a big fan of midjourney magazine and what it represents.

It's true that Midjourney magazine doesn't compete as an art publication in a traditional sense.

But it does something far more interesting: it makes me, a mere mortal, feel like an artist.

Every artwork is accompanied by its prompt, and the prompts work. It's fun to type these prompts into the midjourney discord and see something amazing created.

I can play with certain words, add phrases, modify descriptions and quickly iterate towards something that feels unique and, curiously, "my own".

Sometimes the prompts are surprising, like this one:

"don’t look at the eloquent red circle, surreal, glistening highlights –ar2:3 –s 33": https://shorturl.at/aBMRS

It creates dramatic red-toned images featuring with a haunting surrealistic robed humanoid beneath an ominous orb.

I iterated on this a bit, to explore the impact of different words.

For example, "the eloquent circle has much to say ": https://shorturl.at/nxMO9

Midjourney magazine is a manual for visual creation with AI that even me, a programmer with zero artistic ability, can use to feel like an artist. The prompts are amazing, surprising, and, unlike any other art publication, interactive.

Rendition creates a two-way sync between design and code. It’s a figma+github plugin that will write UI code for you from designs. To do this, it compiles figma designs into react code and creates a pull request in your repo.

It also syncs designs from code back into figma, and this lets designers more easily work off of the latest designs (“pull latest designs from code”). And because Rendition syncs design and code in both directions, iterating on UI works well by generating nice diffs for PR review and allowing you to edit the generated code.

We think of it as an extra pair of hands to implement html/css for us. It’s best for relatively small product teams who need to get a lot done quickly.

Let us know what you think, and yell at us if you think we’re wrong

We hope to carry the torch forward from PageDraw’s efforts in this direction: https://news.ycombinator.com/item?id=28201339 https://news.ycombinator.com/item?id=19273403

Interesting, do you have experience with web agency work? I've never thought about the challenges they face, but they must face this challenge all the time

Ah, so what you're saying is that a big problem that dev+design teams will encounter is that they need to be able to iterate on designs together, so a real solution here would need to be able to integrate updates to the design into the codebase (and possibly vice versa). Am I understanding your point correctly?

Cool to hear about this

I appreciate the thoughtful response.

Wiring up a frontend to a backend isn't a trivial task, but is pretty typical software engineering. If part of the work (the ui) could be done for you, wouldn't that be a clear value-add?

Sounds cool, lmk if you pursue it.