AI coding tools are evolving incredibly fast - but the user experience may not be keeping up. In this episode, Matt shares his first experience trying Codex on Windows and how a simple attempt to generate a classic Snake game quickly turned into a confusing experience filled with permission prompts, unclear setup steps, and rapidly draining usage credits. This sparks a larger discussion about whether AI development tools are moving so quickly that UX is being left behind. In this episode Matt and Mike discuss the gap between tools like ChatGPT and more advanced coding environments like Codex, why developer tools can still feel intimidating even with AI doing the coding, and how today’s AI ecosystem feels a lot like the early days of crypto - powerful but sometimes chaotic.
A growing sentiment on this show is that AI UX is being left behind in the name of constant innovation. This situation hit a critical point with us when we started seeing the UX of AI developer tools such as Codex (main focus of our discussion), Claude Code, and more.
Matt recently installed Codex for Windows to try it out.
Instead of a guided experience, the tool immediately opened a file dialog asking him to select a folder - without any explanation of what it was for.
After skipping that step, Codex presented a playground environment with example prompts like:
Matt tried the Snake game example.
But instead of simply generating code, Codex immediately began:
npm testAll while the user has very little context for what's happening.
One thing that stood out immediately was usage metering.
While experimenting with Codex, the system was already consuming 5% of available usage during a simple test.
For users experimenting with the tool for the first time, this creates anxiety:
This is very different from tools like ChatGPT, where experimentation feels relatively safe.
ChatGPT succeeds because the interface is extremely simple.
You open the app and see one thing:
A prompt box.
From there, users naturally learn through conversation.
They can ask:
The UX encourages exploration.
Codex, on the other hand, feels like it drops users directly into a developer workstation environment.
Without onboarding, that environment can feel intimidating - even for experienced developers.
A common narrative in AI right now is that AI is democratizing coding.
We often hear that anyone can now build apps without developers.
But tools like Codex highlight an interesting reality:
Even with AI doing the coding, the surrounding tooling still assumes familiarity with:
For someone coming from a ChatGPT-only experience, that leap can be massive.
In contrast to Codex, many modern apps are obsessed with onboarding.
Some apps go so far as to guide users step-by-step through every feature.
Matt compares this with software like QuickBooks Online, which frequently displays tutorials, feature announcements, and walkthroughs.
While this can sometimes feel excessive, it does highlight the opposite problem:
Codex offers almost no onboarding at all.
The result is a tool that feels powerful but also somewhat opaque.
This situation is reminiscent of the early Web3 and cryptocurrency boom.
At that time, there was enormous hype around blockchain technology, but the user experience outside of centralized exchanges was often confusing and risky.
For example:
In many ways, today's AI tooling ecosystem feels similar.
The capabilities are impressive, but the UX maturity isn't always there yet.
One possibility is that this is simply a first-generation problem.
AI capabilities are evolving so quickly that developers are prioritizing:
UX refinement may simply be lagging behind the technology.
But if AI is truly meant to democratize development, eventually these tools will need:
Prices subject to change and are listed in USD
Learn to code using Scrimba with their interactive follow-along code editor.
Join their exclusive discord communities and network to find your first job!
Use our affiliate link for a 20% discount!!
We receive a monetary kickback if you use our affiliate link and make a purchase.
