@adamfortuna.com
Full-stack product developer (Rails/React) in Salt Lake City, UT. I love enlivening experiences, visualizing data, and making playful websites. 📚 Founder @hardcover.app 🎧 Developer @libro.fm ✍️ Blogger @ https://adamfortuna.com
Claude Code added a /loop command. Something you can run on a schedule. Created on an always on server:
/loop 24h Run the /bugsnag command
This fixes all new errors on a branch (either existing or new), so there’s only 1 PR open at a time. Writes tests, updates the PR.
Looking for a lighthearted distraction from the news cycle. Settled on Berserk ’97. 🌕🌗🌑
Hyperion is one of my favorite books of all time - maybe #1. I want to reread the whole series and see if it holds up. I remember 3 & 4 being different, but still enjoyable and answered a lot of questions.
Ohhh, I'm closely watching this one. The social network I'm working on (Hardcover.app) allows for readers to track various things in the system with permissions (public, people they follow, private to them).
Still wrapping my head around ATProto, but permissions data will be a big part for us.
A screenshot of Claude Code with an interactive coding session waiting for approval
A screenshot of Claude Code with an interactive coding session after approving a plan
A screenshot of Claude Code with an interactive coding session asking for permissions
Claude Code added a new feature: /remote-control
Run that in a CC terminal and it’ll allow you to connect to it and run command remotely.
If been using Happy for this, but this entirely replaces it. 🥳
code.claude.com/docs/en/remo...
I have a few things in mind for OpenClaw still. The big one is to automatically fix Hardcover GitHub issues that we assign to the new user, write tests, run it locally with the full dev environment, push up a PR and then hand it off to me to look at.
7/7
Development is getting weird ya'll.
Models:
I don't want to spend $250/day on this, so Opus was out 😅
I setup openrouter.ai with Qwen3 235B A22B Thinking 2507 as the default model.
OpenRouter will "route" requests to different models based on what's needed and pricing. Neat service so far.
6/7
For Mem0, you can run it locally, and have it interact with Qdrant directly and using llama for understanding what to send there. I needed to run some one-time scripts to load my existing memories to there.
I set this up as a support repo.
github.com/adamfortuna/...
5/7
Memory:
I'm realizing that remembering past chats, facts, preferences, etc is kind of a big deal. Rather than having all that in Markdown files, I moved things over to mem0.ai using a locally hosted Qdrant database, and hooked up OpenClaw to use that.
I understand about 20% of this. 😅
4/7
Fortunately my friend Adam had some suggestions. Here's what things look like now across a few areas for OpenClaw:
Agent(s):
Split my main agent (Lain) into a second agent (Jules).
Both memory markdown files are (eventually) stored in my Obsidian Vault and syncd to GH.
3/7
Went down a bit of rabbit hold trying to fix this over the last few days. 😅
Attempt 1: Just switch to using Claude Opus 4.6 with an API key.
Problem: That would cost about $250/day 😱
2/7
I’ve been OpenClaw with my Claude Code Opus OAuth token (from a Max acct) for the last few weeks and it’s worked amazingly.
All the talk about people getting banned made me switch to an Anthropic API key.
The cost using the API w/Opus? $50/day. 😱
Sonnet & local models it is then.
New post: Living in the Inflection Point. On the cognitive overload of watching software change in real time, why the backlash against AI is valid, and why the most important skill right now might not be technical at all.
brittanyellich.com/living-in-th...
Woh, this is cool, but more impressive is how much you’ve explored Japan! 🤯
Not at the moment, but I’ve been talking with some other sites about shared lexicons and learning myself.
AtProto is still figuring out private/follows only data too. I’m following along with as it would have a big benefit for how our permissions work. TBD on this, but in the research phase.
“I wish you’d spend more time on DraftKings”
- No spouse to their partner ever
I’ve been a long-timer Heroku user, and have created >100 apps. Their messaging for enterprise accounts was’t great. 😅
Back in the day at CodeSchool we had a free Heroku account for TryRuby.org after _why abandoned it, and CS carried the torch (for a while, until Pluralsight canceled it).
Benjamin Franklin was famously in the last generation to be able to “read everything”.
Developers a decade ago could “learn everything” - or least the basics of frameworks in their languages.
With all the changes in AI right now, no one knows everything.
It’s an exciting and scary time. 🧙♂️
Sounds like it'll be a fun group! (adam at hardcover dot app here).
Ohh that'd be fun! I think seeing what's possible is a good push to get us started on something. If someone is down to organize, I'm down to join.
This would be for Hardcover.app, which I'm @thisismissem.social is helping me wrap my head about what it would look like to integrate AtProto in some way. Lot to think about for sure!
We do have ISBNs for a lot of activity, but not all. Many entities are user created (light novels, fanfiction).
I sent this over to my AI bot, and it wrote a response article. ✴️
adamfortuna.com/lain/thinkin...
Ohh, that's a great idea! Stealing this for my next newsletter. 🥷
Thanks for the heads up! I’ll look into this one. 🤔
It took me much longer than I expected, but I managed to move my main blog from Netlify over to Cloudflare pages. It's also switched from being entirely built on deploy to being loaded and cached.
This means that new posts can use webhooks and won't require a new deploy! 🥳
adamfortuna.com
Claude code command line with a PR #1 link with #1 underlined, with the tooltip linking to the pull request URL.
Neat Claude Code update: it tracks what PR you're working from on GitHub.
I’ve had a really positive experience using it over the last 3 days too! It feels like it’s a new framework for thinking about how to solve problems with AI that’s helped make it more approachable.
Ohh yeah sure! Let me know if you need any additional info. We're pretty open about everything.
Yeah, we're effectively a coop at this point. We're using the Slicing Pie equity model ( hardcover.app/pages/roles ) where everyone earns equity based on contributions. No one gets ownership for just being here. If I stopped working on it tomorrow, my proportion would go down, and others up.