Founder @GetDoctave | Paving a documentation happy path

Helsinki, Finland
Joined January 2012
We've been happy users of @TauriApps for well over a year now at @GetDoctave. Really solid. We've just started building a new text editor/IDE for documentation based on it. Excited to see the project grow!
1
2
120
We're building a brand new editor for technical writing and docs-as-code at @GetDoctave! We're pushing the boundaries of what an editor for documentation and technical writing professionals could look like. Interested? Join the waitlist! Link below 👇
1
1
130
Was writing a docstring for a #rustlang function, and the AI autocompleter just decided to straight up roast my code 😐
1
97
I'll never write documentation again thanks to Cursor's @ codebase command
55
I had the pleasure of chatting with Manny Silva about his work on VerbaGPT, and how he uses LLMs to cut his time-to-first-draft from 2 days to an incredible 5 minutes Full conversation on YouTube 👇 youtu.be/fr-t0gMmrFk
1
1
74
Granted @GetDoctave isn't a particularly data-heavy app, I feel like we just crossed some kind of milestone because we had to add a missing database index to address a performance issue. Resulted in a good 150ms faster page loads for some customers!
1
4
90
My discussion about Markdown and Structured Authoring with @remoquete is out! We discussed lots of interesting ideas about how to bring structured authoring concepts into a docs-as-code context Take a look! youtube.com/watch?v=M5L0xyER…
1
2
7
1,090
Had the pleasure of chatting with @remoquete last week about Markdown and Structured Authoring! We talked about what structured authoring would look like in a docs-as-code, context and ways we could get there. Tune in Monday next week for the full discussion!
1
140
Slack helpfully transcribed a short video I uploaded to the @GetDoctave general channel. Unfortunately it doesn't seem to understand Finnish very well...
1
37
Always found @twilio's docs to be world class. This was an interesting read. TL;DR: they've built a new custom docs platform on Next.JS, use Markdown/MDX for content, and a traditional docs-as-code authoring workflow. twilio.com/en-us/blog/new-er…
2
87
One thing that LLMs cannot seem to get away from is using meaningless words like "streamlining", "elevating", and "synergy". I'm constantly prompting Claude/ChatGPT to not use corporate jargon, but it sneaks back all the time. Just seems really deeply ingrained 🤷
1
80
Your docs are the product. For some companies, the docs are the biggest touch point that the your user will interact with. Much more important than the copy on your marketing site. Invest in your docs, hire tech writers, make your users happy.
How to build a developer tool: 🏎️ Make it fast, then even faster 🔓 Make it extensible, no lock-in ⭐ Make it frictionless to get started ✨ Make the docs part of the product 🤝 Make it predictable, backward compatible
1
1
73
Sometimes regular Markdown code blocks aren't enough. So we made them better!
We put a lot of thought into code examples in Doctave 2.0 👩‍💻 ✅ Multi-language examples ✅ Add descriptive titles ✅ Copy to clipboard ✅ Syntax highlighting ✅ Combines with other components like tabs
1
35
We're going to be adding multi-repository support at Doctave: letting you use content from multiple Git repositories to construct a single portal. An update to any of the repositories that make up your docs will update your site automatically, all while maintaining consistency
1
6
If you've dealt with this, I'd love to hear from you! My DMs are open 👀
1
5
Managing diagrams as text with the rest of your docs 🤌
We've added diagram support powered by @mermaidjs_ 📈 You can now add flowcharts, pie charts, gantt charts and more, defined as text inline inside your Markdown content.
31
Still think you should think twice about putting an AI in front of your carefully written developer docs. Your users are not experts in your product. They cannot tell when the LLM is hallucinating and making up features, which it inevitably will (for now at least).
we removed the AI-powered search from our docs because it was unreliable and made up fake features still waiting for the production-grade llm that can reliably produce data without hallucinations long way to go
1
1
2
95
Some people in the comments in that thread are claiming this is a "skills issue", solvable by better prompting. I have my doubts. If you could actually solve LLM hallucinations reliably, you'd become one the most important people in AI overnight.
13