Skip to content

Comment on We don't use AI in any of our design or production processesparent

Comments

Summarising emails that then misses the crucial information is not saving any time.

Ironically due to an overloaded inbox these days, I've been using a LLM as the world's most inefficient e-mail filter.

It does a decent job at this, and gives at least a nice starting point in the morning for a prioritized hitlist to power through. Gets most things pretty accurate, but I only use it to sense "importance" and not much else.

Certainly not life changing, just a nice to have.

I've had far more real gains with home IT projects. Tons of home automation integrations and "silly" projects I've been wanting to get done for a decade but never really had the energy to want to deal with. These sorts of things also allowed me some time/excuse to play around with agents to see what the fuss was about.

I now have the world's most complex garden irrigation/sensor setup complete with automated watering and grafana panels! I could have tied everything from the hardware through software together myself, but I just was not interested in installing the 856th Ubuntu VM in my life, configuring InfluxDB from scratch, etc.

I can see the future potential due to the above, but I also see how it went off the rails easily and also produced some laughably inefficient/unmaintable stuff like Flux queries which toss what should be variables as statics directly into the grafana panels, etc. Add a single new sensor to a group and you'll be re-writing the panel queries. Just horrible ground level design only maintainable via brute force. But when you just tell it to add the thing and it does... I can see why many feel it's a superpower.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.