I think “parking traffic” might also be an issue.
Like you go to eat out in the city and there isn’t any parking, you just have the car circle the block for an hour.
Enough cars circling the block, and it’s gridlock.
In the end, there will be regulation around this. Similar to the no idling zones. There will have to be “no idle circling” around the block.
I kind of wish there was a better way to automate this.
I schedule the recurring takeouts. But this just sends me an email when it’s ready. I have 100s of gb, which means lots of 50gb zips.
I them need to log onto my server and manually pull them down.
The anying thing is that they expire pretty quickly. More often then not, I miss the window, and then need to re-request it and wait a few days. Then I miss the window again…
I can probably hack something together, but wish there was something that just worked.
One of the things I like doing for large tasks is that once we are aligned on the plan, I’ll tell it to work on each phase. Once the phase is completed, commit the changes and ask codex for an adversarial review. It should then work with codex to resolve any issues that arise. Once the phase is done, move onto the next one.
I add that any questions that arise, it should work with codex to make a call. I’m the final summary it should enumerate the list of questions, options and decision so I can make the final call.
The code produced by this flow is really nice and well done.
I keep going back and fourth on this.
The two thing I like about talking is that I’m brainstorming sessions I can just do a brain dump of my thoughts.
Another thing I’ve come to like is that it gives my hands a rest. I’m constantly inputting into the AI (I use @conductor_build) and after several hours my wrists start to hurt.
The annoying thing about talking is that none are perfect. For brain dumps it’s fine and the ai figured it out from context. For detailed messages and personal communication, I almost always have to edit something.
At this point all I want is to be able to have an LLM in the loop. Before a notification is delivered it should run through an LLM filter, which I can prompt. So I can describe exactly what I want:
- Instant delivery with a ping
- silently deliver
- add to daily summary
- reject/block
It would also be nice for the system to manage the prompt as well. So if I get such a notification, I can tell it not to show me these again, explain what I don’t like about it, and it can learn for the future.
@GergelyOrosz It’s probably common enough where large companies that use Jira acquire smaller companies that use Linear.
But I don’t think they would consider replacing everything with Notion. Jira is pretty entrenched and hard to move away from.
The way I see it, these comparison pages are more about SEO. You kind of want to try getting a web presents for people searching for comparisons.
I’ve actually found it useful for finding alternatives to the product I know. So say I want to see what things are similar, I look at their comparison pages to find other options (which is obviously not what they intended 🤷♂️)
Once we are asking for features, I’d love to have a “copy thread as markdown” button.
Though not sure how to handle attachments cause those are needed for context as well.
Maybe if there are attachments, offer a “download zip file” which includes a markdown file of the conversation and all the attachments included.
The call screening on iOS 26 is really nice. However, it still passes the call on to me. It would be nice if I can mark calls as spam, it can them intelligently (using an LLM) decide based on the screening conversation.
This is a spam call I get at least once a day and always from different numbers.
Would be nice if I can tell it to always ignore any calls about taxes from unknown numbers.
At this point all I want is to be able to have an LLM in the loop. Before a notification is delivered it should run through an LLM filter, which I can prompt. So I can describe exactly what I want:
- Instant delivery with a ping
- silently deliver
- add to daily summary
- reject/block
It would also be nice for the system to manage the prompt as well. So if I get such a notification, I can tell it not to show me these again, explain what I don’t like about it, and it can learn for the future.
@CoinbaseDev is this legit? no subject and no attachment, and asking for an NDA just to see what pricing you're going to try and start charging us in 2 weeks...? what's going on here?
The divide between people who say coding agents do all the work really well and those who say AI doesn't work is fascinating.
I think part of the difference comes down to the type of person. "Builders" tend to love AI because it lets them create far more than they could before. People who genuinely enjoy the craft and challenge of writing code are often more skeptical because they're evaluating AI on a different axis.
I mean this less as an observation about code quality and more about the inherent biases people bring to the table. If your favorite part of software development is solving problems by writing code, you're naturally less excited about a tool that does more of that work for you.
Malware increasingly “lives off the land”. Abusing trusted, preinstalled tools to blend in and evade detection.
As AI dev tools like Claude Code and Codex become widespread, expect attackers to start leveraging them too.