The two meanings of unacceptable

Lucas Montano posted that it's unacceptable for a professional dev not to be able to pay for one or two AI subscriptions. The internet caught fire. Part of the outrage makes sense. Part of it is people talking past each other.

What bothered me wasn't the post itself. It was seeing two meanings of "unacceptable" glued together as if they were the same thing.

Where people got lost

There are developers earning R$ 1,500 a month. If that's you, the unacceptable part isn't the Claude or Codex subscription. The unacceptable part is the salary. And that's not your fault.

Mixing the two turns into a crooked moral judgment, as if someone living on the edge had an obligation to squeeze their budget even harder to "stay up to date". No. Fair pay comes first. Tools come later, when they fit.

Now, if you can afford it (or can find a way without wrecking your finances) and you simply don't use AI day to day, then "unacceptable" changes target. This isn't influencer moralizing. It's the same kind of lag we've already seen with English: the market doesn't wait for you to accept the tool.

Do the math on your hour

Before arguing about whether the company should pay (spoiler: in many cases it should), it's worth looking at the cold number.

What is one hour of your time worth? Take your monthly take-home pay and divide it by the hours you actually work. Not the fantasy of 160 hours. What's left after meetings, context switching and interruptions.

If one extra hour of output a month already covers a Codex Pro, the subscription stops being "spending on hype" and becomes arithmetic. Being more productive isn't typing faster to impress your boss. It's producing more value per hour: the same freelance gig, the same sprint, with less friction between the idea and code that's ready for review.

What if the company won't pay (or won't allow it)?

Ideally the company pays. You deliver more, it gains. The math works on both sides.

When it doesn't pay, or when it blocks AI on the project, there are still options. Pay out of pocket for personal and study work. Use it where the policy allows. Move somewhere else if the restriction is dogmatic and permanent. The point isn't to romanticize "investing in yourself". It's to recognize that if the brake is purely cultural, the one standing still is you, not the market.

Some people treat the subscription as taking money out of the budget to pay for a tool. I treat it as a goal: find a way to pay for it. One less hour of distraction a week. A short freelance job. A model with subsidized tokens when the full-price API doesn't fit. Chinese models come into this conversation for the same reason: a cost per token that makes experimenting viable.

Why "not using it" got expensive in 2026

I say this being conservative: with Claude and Codex, using subsidized tokens instead of insisting on the most expensive path, I can deliver an order of magnitude more than I did without this kit. In the video I land on roughly 20x. The exact number matters less than the direction.

AI also frees up time to study and to pull things out of the drawer that had been sitting there. It's not just "closing tickets faster". It's having bandwidth left over to learn what the next ticket is going to demand.

If you can't save one hour a week with the tool, or don't try hard enough to open up that hour, the conversation shifts from "I don't have the money" to "I'm not testing the gain". Those are different problems. Mixing them is what set Lucas's post on fire.

In short

The post hit a nerve and the thread scrambled the diagnosis. A dev earning R$ 1,500: the problem is pay. A dev with room in the budget who refuses to use the current kit: the problem is falling behind by choice, in the same spirit as ignoring English when the job already requires it.

Do the math on your hour. See whether one extra hour a month pays for the tool. Push the company to cover it. If it won't, treat paying for it as a goal, not as guilt. And use the gain to study and unblock what was stuck, not just to type more lines.

What's unacceptable, in the good sense of the word, is pretending 2026 is still the same game without this leverage.

Watch the full video at It's unacceptable for a dev not to use AI in 2026 (in Portuguese). There I separate the two meanings of "unacceptable", do the math on the hour, and talk about Codex Pro, subsidized tokens and why this looks a lot like not speaking English.