Yeah, if you're used to writing tickets for engineers your choice now becomes 'do I hand this off to a human or an AI' for some of the tasks – and that can be fun too
13.01.2026 13:09
👍 0
🔁 0
💬 0
📌 0
Yeah, if you're used to writing tickets for engineers your choice now becomes 'do I hand this off to a human or an AI' for some of the tasks – and that can be fun too
It gets interesting when a lot of context across different parts of a system is required. We usually hold these in our heads when developing, but very subconsciously. Giving the LLM all the right context is still hard.
You haven't missed much and are spot on :)
A lot of hype these days, but other parts of coding with AI is actually *a lot* easier than doing it by hand. I think I'll never again code a Slack integration for example, just because it's so simple for an LLM to one-shot these known patterns (OAuth flow)
Interesting change happening at work, where reviewing all the code that's so easy to produce these days is apparently becoming the real challenge.
buildsharerepeat.substack.com/p/three-llms...
New question I ask myself in code reviews: is my feedback going to help this person grow, or just generate a better prompt?
Finally recording a new ep for the pod right now ⚡️
Still sick and voice won't allow a podacst ep this weekend. On a positive note, Linear integration is working. Even two-way, so Feedbackview sends back a notice to the user when a ticket is closed ⚡️
Started working on a Linear integration, to automatically create tickets from incoming user feedback and bug reports, based on some config rules 🔥
Yeah, true. Going to be interesting to find the right balance of what feedback should open a ticket at all :)
4. PR gets merged
5. Linear closes ticket automatically
6. Feedbackview notifies customer autoamtically
The automation rules would let users define what should actually auto-create a ticket, eg. only bugs that have a high impact score.
linear.app/changelog/20...
Oh yeah, Linear and GitHub integrations are definitely coming! Trying to grasp what makes most sense to build first at the moment.
It could work like this:
1. Feedback received in Feedbackview
2. Automation rule creates Linear issue automatically
3. Linear hands off to GitHub Co-pilot automatically
New podcast episode recorded, contacts feature going hard already. I'm having a blast over here 🤘
Toddler asleep, time to build the next table stakes feature – contacts 📇
Thanks Veni! 🙌
If bootstrapping podcasts are you thing, I’m getting back into it! Trailer Episode just launched! 🎙️ buildsharerepeat.com
Showed the kiddo how to repeat a song on his player today, big mistake.
Toddler is napping, time to ship another table stakes feature: Internal notes 📝
True – gotta tripe check all the things, but still much faster!
AI is definitely a force multiplier these days for me, especially when working solo on a project. Still took a good 2 months for me, working on it on the side. Bet Val spent a good amount of time on the new thing too!
No pressure! :D
Banger domain for a podcast has been acquired 🎙️
Time for some good old weekly build-a-saas updates
Got Notion Mail yesterday and it's pretty neat tbh. I like the simplicity and it's starting to feel like a whole ecosystem 🔥
Not sure if you're into AI IDEs, but if you're using Windsurf, they have an option that lets you reference all the code etc. when you turn "write mode" on, but it can't do any edits. Pretty useful if you want to reason about something without the actual implementation.
Claude 3.7 💯
That sounds unpleasant. You should totally do that :D
Ah that’s even harder to prevent then, oof :-/
It’s like grocery shopping with a toddler. You let them roam and it works okay and the second you become complacent they trash the entire aisle. Always keep an eye on the toddlers 😄
Adds friction, but you could require account activation through email on signup. Would only do if this is a big issue
I always had one Paddle account per product and they were tranferred fully when selling.
