Ship it
"Ship it. 🚀"
This is the final list of tokens generated by Fable ahead of Linkraider going public.
I started working on Linkraider a few years ago, while brainstorming ideas for some personal projects to work on during my free time. I always struggle to keep my favorites bar tidy and there is not enough room there to hold all my bookmarks.
Since I couldn't find any tool fitting my needs, I started to think about how I would design for myself. At that time I just had a three-pages requirements document in google docs and a couple of mockups in Figma, heavily inspired by social platforms like "Twitter".
This draft has been sitting in my Google Drive for more than three years. Then, this year, I started playing with coding agents, soon after Claude Opus 4.6 and GPT 5.4 were released.
At the beginning I just shared the drafted requirements and some screenshots of the mockups I designed years ago, asking Claude Code to design a possible architecture of a web application implementing my idea.
I was then shared a 12 pages architectural document describing all the details and nuances of the different layers of the architecture, from the DB model, the backend layer, the frontend and more.
I asked my new friend to breakdown all the architecture in an implementation plan and, for each step of this plan, to share with me the prompts to describe a coding agent what to do.
That is where it all started. Iteration after iteration. I had a prototype ready in one week, and the more I used the more I had ideas on finetuning, changes, small enhancements, that were all turned into real working code in a matter of minutes.
What happened next....
I got a domain (a few ones, because finding a free domain name was the hardest part of this all) and went through all the steps needed to turn this into something that could be used also by other people with no risks.
I asked the model to do the work form me. Connect to the server, start a Kubernetes cluster, harden it, deploy the nodes required by the architecture. I discovered what skills are, and how to encode all this information in my personal devops team markdown file.
Several iterations were spent on the security of the application, which I will not detail for obvious reasons, and on matching all the GDPR regulations needed for a platform like linkraider.
Iteration after iteration. Yesterday Anthropic made available to everyone Fable 5. A finetuned version of Mythos, only for a couple of weeks. I couldn't find a better time to finish the project.
The last performance test confirmed I can sustain with my small cheap server a workload of more or less 1000 users.
I have issued the last prompt, just before writing this post.
> "/devops deploy in prod and confirm we are all good to go live."
---------------------------------------------------------------------------------
Ship it. 🚀
※ recap: Pre-go-live prep for Linkraider is finished: security review, hardening, backups, OS updates, monitoring, and final deploy all verified live in production. Nothing pending; next action is your actual product launch.
---------------------------------------------------------------------------------
Welcome http://www.linkraider.com/
If you are wandering if an LLM was used to write this post, no. I believe our brains can still deliver better tokens than any model.
If used properly.