Is it all just vapourware?

It seems increasingly likely that much of the new "AI" and "agent" software is merely vaporware. The author has been trying to embrace the concept of "agentic development" but encounters an overwhelming amount of friction within the ecosystem. Engineers seem to have free access to unlimited tokens, but lack systematic review and quality assurance processes for usability testing or real-world testing.

This results in ubiquitous code but little regard for user experience. One example given is the ONA platform, which promises to automate the refinement of feature requests into actionable items throughout the development process. However, the author's first interaction with the platform was a failed login attempt on the desktop version, followed by a cumbersome and ineffective onboarding experience.

Even after getting past the login, the platform consumed nearly all of the author's "ONA compute units" trying to integrate with the author's project, adding unnecessary friction. The author is skeptical that AI can actually streamline the software development process, and is turned off by the excessive integrations, authentication requirements, and billing complexities.

The marketing surrounding these products is also overly optimistic. The author believes ONA has potential, but wants a working version that doesn't waste their time and money.