AI changed IT
All paths

Path for working developers

The code writes itself. Now what?

Four stages from using an assistant with care to judging whether an AI feature is good enough, and an optional fifth toward forward deployed work.

  • 5 stages
  • 164 hours
  • 21 weeks

Your pace, hours a week

5 stages, 164 hours

0%

Stage 1, use an assistant with checks

Trust it less, test it more

  • 24 hours

This week

Take the last thing AI wrote for you. Explain each line out loud, then write one test that would fail if it were wrong.

More free options
Paid options

Ranked by fit, ratings and freshness. We earn nothing from these links.

You end up with

One small change to a real project, with tests and a short note on what AI got wrong.

More about this path

Where this gets you

Can ship a small tested app with an LLM feature, explain every part of it, write an eval set, and check it against the main LLM security risks.

Just lost your job?

If you have just lost your job, do this first

Stage 1 only, about 2 to 3 weeks at around 5 hours a week. Do the Google module and the first Zoomcamp module, then make one small change to a real or personal project with tests and a short note on what AI got wrong. Do not pause your job search to study. Do not buy an exam at this point.

What you need

Basic programming, the command line and Git. An AI coding tool: GitHub Copilot Free costs nothing and needs no card, or you can run local models with Ollama. The main course in stage 2 expects about $10 to $15 of tool and API spend.

Worth the fee, or not
CertificateFeeValidVerdict
AWS Certified Generative AI Developer, Professional$3003 yearsFor people focused on AWS. Recommends 2 or more years building.
Google Professional ML Engineer$2002 years on renewalML engineers on Google Cloud only.
Microsoft Azure AI-103Varies by country, not shownAnnual, free renewalAzure roles only.
GitHub Copilot GH-300Varies by country, not shown2 yearsOne product. Little lasting value.
NVIDIA NCA-GENL$1252 yearsNiche. No hiring evidence found.

Before paying, count how often the exact name appears in 10 to 20 postings for roles you want. Do not pay for any of these after a job loss unless those postings name it.

What the data says
  • US software postings are up almost 15% since February 2025 and still about 27.5% below February 2020. Senior roles were 71% of the gain from May 2025 to May 2026.

    Indeed Hiring Lab, 8 Jul 2026. Moderate to strong for postings. Ads are not hires.

  • US tech firms announced 171,501 cuts in 2026 to October. Across all jobs, employers cited AI in 121,401 announced cuts. We found no source that counts how many developers lost work because of AI.

    Challenger, Gray and Christmas, 5 Nov 2026. Moderate: announcements, US only.

  • Nobody has a reliable current figure for how much AI speeds up experienced developers. The best recent trial says its own numbers are very weak.

    METR, 24 Feb 2026. Moderate to weak: randomized, damaged by who opted out.

  • Postings for forward deployed engineers, who build with AI inside a customer’s team, rose more than 1,000% from January to August 2026 against a year earlier. All tech postings grew 13%.

    Fortune, from Lightcast data, 3 Sep 2026. Moderate for the rise in ads. Ads are not hires, and a fast rise from a small number is still a small number.

What this path cannot do
  • No exercise here has been tested on experienced developers over a month.
  • Review queues, usage mandates and raised targets come from team rules. Study alone will not change them.
  • Finishing this path builds skill and gives you things to talk about. It cannot promise work.

Checked 21 November 2026.