The AI Skill Anxiety That Only $20/Month Developers Can Afford to Have
There's a fascinating discussion happening right now on Dev.to.
Developers are asking: Is AI making me worse at coding?
The most-reacted article this week is 'I Used to Love Coding. Now I Just Prompt.' It's struck a nerve. Developers are worried about skill atrophy, loss of deep thinking, the slow erosion of the ability to hold complex problems in their heads.
This is a real concern. I've felt it too.
But here's the thing nobody in that conversation is saying:
This is a luxury problem.
Who gets to worry about AI skill atrophy?
Developers who worry about whether AI is making them lazy are, by definition, developers who are using AI enough to form an opinion.
That requires $20/month for ChatGPT Plus. Or $20/month for Claude Pro.
Let's be concrete about what that means globally:
- Nigeria: $20/month = 3-4 days of average developer salary
- Philippines: $20/month = 2.5 days of typical freelance billing
- Indonesia: $20/month = Rp320,000 — a significant weekly grocery budget
- Kenya: $20/month = KSh2,600 — nearly a week of transportation costs
- India: $20/month = Rs1,600 — 10x what a local SaaS subscription costs
For developers in these markets, the skill-atrophy debate is academic. They're not using AI enough to atrophy. They're rationing it.
The rationing problem is worse than the atrophy problem
Here's what rationing looks like in practice:
// Developer who can afford $20/month AI:
// "Let me ask Claude to review this function"
// Gets review, iterates, learns the pattern, ships better code
// Developer rationing expensive AI:
// "Do I really need to use an API call for this?"
// Decides it's not worth the mental accounting
// Solves it manually, correctly, but slower
// OR: asks a vague question to get maximum value per query
// Gets a vague answer, doesn't learn the pattern
The developer who rations is not free from AI's influence on their thinking. They're just getting the worst of both worlds:
- They're not building fluency with AI tools (which matter for employability now)
- They're spending cognitive budget on cost-consciousness instead of the actual problem
- They're not iterating — one shot per session, hoping for the best
The research on AI-assisted coding found developers felt 20% faster but measured 19% slower. That gap exists partly because developers changed their behavior when AI was present. They started writing differently — more skeleton code, less thought about edge cases upfront.
But that behavior change requires trust in the AI. Trust requires enough usage to form habits. And enough usage requires affordability.
The experiment you can't run if you're rationing
The developers writing thoughtful articles about AI's effect on their skills have had the chance to actually experiment.
They've tried different prompting styles. They've had Claude do the whole thing and then done it themselves to compare. They've noticed patterns: "I reach for AI faster on Fridays." "I think less carefully about naming when I'm going to ask AI to rename anyway."
This is valuable data. But you can only collect it by actually using the tool.
At $20/month, many developers run the experiment for two weeks, see the bill, and stop.
At Rs165/month ($2), Rp32,000/month, ₱112/month — you can run the experiment for a year.
The access gap creates a skill gap
Here's the long-term problem:
Developers who can afford to experiment with AI extensively are learning:
- When to trust AI output (and when not to)
- How to decompose problems for AI (a genuinely new skill)
- How to write prompts that get consistent results
- When AI makes them faster vs. when it's just noise
Developers who can't afford to experiment don't learn these things.
In 5 years, "AI-native developer" will mean something real. It'll be a skill cluster. And like most skill clusters in tech, early access to the tools during the learning window matters a lot.
The conclusion the debate is missing
Is AI making you a worse coder?
Maybe. If you use it as an autocomplete machine and stop thinking about the problem yourself.
Maybe not. If you use it as a thinking partner and push back when it's wrong.
But the more important question:
Are you getting to find out?
Or are you doing a cost-benefit calculation every time you open the chat window?
The skill-atrophy debate is worth having. But it's only worth having if you have enough AI access to actually form an opinion based on evidence.
For developers in markets where $20/month is a significant daily-wage multiple, the debate starts at affordability.
SimplyLouie offers Claude at Rs165/month for Indian developers, Rp32,000/month for Indonesian developers, ₱112/month for Filipino developers — roughly 10% of the standard ChatGPT price. simplylouie.com