Vibe coding platform Anything arrives, our hands-on suggests caution
(2025/08/14)
- Reference: 1755180010
- News link: https://www.theregister.co.uk/2025/08/14/anything_vibe_coding_platform_released/
- Source link:
Hands On Create has declared its vibe coding platform, now called Anything, production-ready at version 1.0, with support for both web and mobile applications – although our quick hands-on generated a host of errors.
The business was founded in 2020 by former Google product managers Marcus Lowe and Dhruv Amin, now joint CEOs. The Reg received a press blurb which is among the most breathless we have seen, including Lowe's statement that "users are going from idea to App Store to first customer in days" and "anyone can build and monetize apps as easily as selling t-shirts."
Create claims 30,000 monthly active users, that it processes more 20,000 projects daily, and that its Frontier AI agent is "so reliable you can build entire apps without looking at the code once, under the hood."
[1]
All hosting is managed by Anything, or the code can be downloaded for using elsewhere. [2]Plans start from free to $80 per month, based on credits per month, where credits are somewhat opaque but related to AI interactions.
[3]
[4]
The idea is compelling but we like to try things out. The Reg signed on to Anything with a trial upgrade to a Pro plan. No password is needed as signing in is via magic email links – a method which delegates security to the email system. Never mind, we got started using the same prompt we have used to try out Amazon's [5]Kiro and Google's Firebase Studio. The application is for managing a rota of volunteers to run game sessions, and the prompt includes a request for features including auto-allocating volunteers based on their submitted availability, and emailing them reminders before sessions.
After thinking for a bit, Anything said that "the system is ready for volunteers to register, set availability, and for administrators to manage the rota." The auto allocation and email features would need "additional development," it said.
[6]
The generated application is JavaScript using React and Next.js, using Neon serverless Postgres for the database. Unfortunately Anything was over-optimistic and the application did not work at all. The log showed hundreds of errors and the preview screen just said "Loading."
Anything says system is ready; preview stuck on Loading
We asked the AI to fix the errors and it said there were "missing API endpoints." It did some more work and said the "volunteer rota system is now ready." But it was not. It is possible to sign up as a volunteer, but once signed up, the page still says "apply as volunteer" and a second attempt says "volunteer application already exists." Volunteers also have access to the admin screen, which should not be possible. The log still shows many errors. It is possible to view the code though many of the files state that "this file was generated by Create. You may edit it but doing so may cause issues in your app."
The new coding reality, perhaps, is to carry on chivvying the AI until the application works more or less as expected, but it is hard to be confident that the end result will not have non-obvious problems that impact security and/or functionality in unexpected ways. It also seemed odd to us that the Frontier AI sails on regardless of logs full of errors, confidently reporting its success to the user.
It is likely that an experienced React developer could easily find and fix issues and that the AI-generated code may speed productivity; yet the stated goal of Anything is to enable apps to be built without looking at the code.
[7]AI model 'personalities' shape the quality of generated code
[8]GitHub CEO: Future devs will not code, they will manage AI
[9]JetBrains previews Kineto for vibe no-coding
[10]Devs are frustrated with AI coding tools that deliver nearly-right solutions
We may be unlucky. Views on X vary: one user [11]said "I published an app in 18 min from idea to execution" but another [12]reported that "I'm in a never-ending loop of paying for what their program keeps screwing up." Another [13]said : "please Anything can you make a proper and robust testing for built apps before launch," something which sounds to us too like a good idea.
Vibe coding platforms are everywhere. Vercel's well-established v0 tool is [14]now called v0.app with new agentic powers, and the company claimed that "anyone can go from idea to deployed app with UI, content, backend and logic included." Tools vendor JetBrains is [15]having a go at its own no-coding, AI-driven offering. AWS has launched [16]Kiro (with which we were more successful), aiming more at assisting than replacing developers.
It is early days and companies are still feeling their way towards what works; in the meantime over-promising appears to be normal and we suggest caution. ®
Get our [17]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://www.createanything.com/pricing
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://devclass.com/2025/07/15/hands-on-with-kiro-the-aws-preview-of-an-agentic-ai-ide-driven-by-specifications/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2025/08/13/ai_model_personalities_shape_the/
[8] https://www.theregister.com/2025/08/07/github_ceo_ai_coding/
[9] https://www.theregister.com/2025/08/05/jetbrains_kineto/
[10] https://www.theregister.com/2025/07/29/coders_are_using_ai_tools/
[11] https://x.com/pavan2254/status/1954796639274627325
[12] https://x.com/SP7371095179209/status/1955435965167886340
[13] https://x.com/LyongaEkema/status/1955002162896921022
[14] https://vercel.com/blog/v0-app
[15] https://www.theregister.com/2025/08/05/jetbrains_kineto/
[16] https://www.theregister.com/2025/08/05/jetbrains_kineto/
[17] https://whitepapers.theregister.com/
The business was founded in 2020 by former Google product managers Marcus Lowe and Dhruv Amin, now joint CEOs. The Reg received a press blurb which is among the most breathless we have seen, including Lowe's statement that "users are going from idea to App Store to first customer in days" and "anyone can build and monetize apps as easily as selling t-shirts."
Create claims 30,000 monthly active users, that it processes more 20,000 projects daily, and that its Frontier AI agent is "so reliable you can build entire apps without looking at the code once, under the hood."
[1]
All hosting is managed by Anything, or the code can be downloaded for using elsewhere. [2]Plans start from free to $80 per month, based on credits per month, where credits are somewhat opaque but related to AI interactions.
[3]
[4]
The idea is compelling but we like to try things out. The Reg signed on to Anything with a trial upgrade to a Pro plan. No password is needed as signing in is via magic email links – a method which delegates security to the email system. Never mind, we got started using the same prompt we have used to try out Amazon's [5]Kiro and Google's Firebase Studio. The application is for managing a rota of volunteers to run game sessions, and the prompt includes a request for features including auto-allocating volunteers based on their submitted availability, and emailing them reminders before sessions.
After thinking for a bit, Anything said that "the system is ready for volunteers to register, set availability, and for administrators to manage the rota." The auto allocation and email features would need "additional development," it said.
[6]
The generated application is JavaScript using React and Next.js, using Neon serverless Postgres for the database. Unfortunately Anything was over-optimistic and the application did not work at all. The log showed hundreds of errors and the preview screen just said "Loading."
Anything says system is ready; preview stuck on Loading
We asked the AI to fix the errors and it said there were "missing API endpoints." It did some more work and said the "volunteer rota system is now ready." But it was not. It is possible to sign up as a volunteer, but once signed up, the page still says "apply as volunteer" and a second attempt says "volunteer application already exists." Volunteers also have access to the admin screen, which should not be possible. The log still shows many errors. It is possible to view the code though many of the files state that "this file was generated by Create. You may edit it but doing so may cause issues in your app."
The new coding reality, perhaps, is to carry on chivvying the AI until the application works more or less as expected, but it is hard to be confident that the end result will not have non-obvious problems that impact security and/or functionality in unexpected ways. It also seemed odd to us that the Frontier AI sails on regardless of logs full of errors, confidently reporting its success to the user.
It is likely that an experienced React developer could easily find and fix issues and that the AI-generated code may speed productivity; yet the stated goal of Anything is to enable apps to be built without looking at the code.
[7]AI model 'personalities' shape the quality of generated code
[8]GitHub CEO: Future devs will not code, they will manage AI
[9]JetBrains previews Kineto for vibe no-coding
[10]Devs are frustrated with AI coding tools that deliver nearly-right solutions
We may be unlucky. Views on X vary: one user [11]said "I published an app in 18 min from idea to execution" but another [12]reported that "I'm in a never-ending loop of paying for what their program keeps screwing up." Another [13]said : "please Anything can you make a proper and robust testing for built apps before launch," something which sounds to us too like a good idea.
Vibe coding platforms are everywhere. Vercel's well-established v0 tool is [14]now called v0.app with new agentic powers, and the company claimed that "anyone can go from idea to deployed app with UI, content, backend and logic included." Tools vendor JetBrains is [15]having a go at its own no-coding, AI-driven offering. AWS has launched [16]Kiro (with which we were more successful), aiming more at assisting than replacing developers.
It is early days and companies are still feeling their way towards what works; in the meantime over-promising appears to be normal and we suggest caution. ®
Get our [17]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://www.createanything.com/pricing
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://devclass.com/2025/07/15/hands-on-with-kiro-the-aws-preview-of-an-agentic-ai-ide-driven-by-specifications/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aJ4IFQjFu5hWFzbG10kZwgAAAAE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2025/08/13/ai_model_personalities_shape_the/
[8] https://www.theregister.com/2025/08/07/github_ceo_ai_coding/
[9] https://www.theregister.com/2025/08/05/jetbrains_kineto/
[10] https://www.theregister.com/2025/07/29/coders_are_using_ai_tools/
[11] https://x.com/pavan2254/status/1954796639274627325
[12] https://x.com/SP7371095179209/status/1955435965167886340
[13] https://x.com/LyongaEkema/status/1955002162896921022
[14] https://vercel.com/blog/v0-app
[15] https://www.theregister.com/2025/08/05/jetbrains_kineto/
[16] https://www.theregister.com/2025/08/05/jetbrains_kineto/
[17] https://whitepapers.theregister.com/
LLMs generate bullshit.
Fonant
LLMs are great if you want bullshit.
[1]https://link.springer.com/article/10.1007/s10676-024-09775-5
[1] https://link.springer.com/article/10.1007/s10676-024-09775-5
so reliable you can build entire apps without looking at the code once, under the hood.
Kane
Nah, we good.
For those of you old scrotes...
Someone Else
Remember (Ir)Rational Rose, and its attempts to generate working code from Booch (and later, UML) static class diagrams? If you do, you might remember that the (C++) code it generated as impenetrable as it was useless. For much the same reason as what can be generated now. This "Vibe Coding" stuff appears to be the next generation of coding-without-writing-code codswallop. With just about the same results. (Although this does appear to be able to generate abject crap much faster -- thank you, Moore's Law!)
What was old is new again!
Reminds me a automatic spelling correction
Claude Yeller
All the fully automatic AI code generation stories remind me of the days when I used automatic spelling corrections during typing on my mobile phone before I got to my senses and disabling them all became the first thing to do after powering up a new phone or app.
Yes, the spelling of my text was flawless with the spelling corrections, always. Newer applications also got the grammar correct.
But the resulting texts were rarely if ever what I wanted to write, and generally career limiting if send uncorrected.
Fast Forward 2025, now we have AI writing syntactically "correct" (for certain values of correct) code that will not do what you want.
And if it by accident the code does seem to do what you asked it to do, it still might terminate your career when shipped due to "undocumented features".
Analysis
" anyone can build and monetize total shit code as easily as selling t-shirts "
This is not even innovative. Decades ago there was UK TV retailer sitcom called "Never mind the quality -- feel the width".