“I believe that to be the best, you have to smell like the best, dress like the best, act like the best. When you throw your trash in the garbage can, it has to be better than anybody else who ever threw their trash in the garbage can”
—Lil Wayne
(h/t @PirateWires)
Named functions are my hill to get moderately injured on.
Other than having a descriptive name and also not being visually mixed with all the other variables, they are MUCH easier to pinpoint when working with a profiler on performance issues.
I learned a while ago that, if an agent doesn't solve a problem in 2-3 tries, I should stop asking it to try more. It doesn't know, that's that. It's not gonna figure it out in any reasonable time and effort.
What typically works great for me is this prompt:
"Let's stop debugging this. I want to start a new LLM session. Write a document with a detailed description of the problem that we're trying to solve. Describe everything you tried so far and why it failed."
Then I hand that to either GPT 5.4 High or Opus 4.6 High, whichever was the one that couldn't solve it originally. And in 90% of the cases, the other LLM one-shots it.
Now I understand the full picture.
The cleanest fix is...
But actually, the real fix simpler...
Actually wait.
The best fix:
Now the real fix.
Actually, let me reconsider.
OK
Key finding:
Wait
I need a hardware device I can physically punch to stop the agentic session.
2026 update:
Unattractive men love extreme sports, wear makeup and watch anime
Unattractive women are messy pessimists who smoke every day, have conservative politics and listen to kpop
This article is AWESOME and it perfectly shows in what kind of terrible state Tahoe (but also the other 26 releases) is.
EVERY Developer at Apple should be forced to read this article. Zero debate.
@ImSh4yy If you just keep backups, then on every request you can spin up a fresh db and load backup data in an available region.
It can’t go down if it wasn’t up to begin with.
@arvidkahl Stories.
People who had some degree of success often falsely think they know exactly what lead them to it but, in reality, they don't. Part skill, part circumstance - but rarely directly applicable to other people.
Went to a restaurant up in the Swiss Alps, open wifi network had the login popup where you can either type in all your personal information, or log in using DISCORD
🏛️ Knowledge Hubs is RAG on steroids, but even more so, a full CRM replacement.
If AI can understand your files and what's important for your business, it should be able to populate a system of record and retrieve hidden unstructured information on the fly. Think of it as an auto-indexing and auto-populating salesforce/hubspot that focuses on the metrics you care about, saving you from manual input.
Hubs self-regulate the depth of a search query based on the complexity of a question with an agent loop - starting from pre-extracted properties to reading through retrieved files for more info.
We expect V7's next data room for due diligence will be fully generated in a hub 😊
Go give it a spin.
Slumdog millionaire, but with a rich guy who gets to answer a $100k question about an endangered marine species by recalling that one time he couldn’t dock a yacht
Slumlord millionare or something
Gonna wait until I can one-shot this
I've been playing with Apple Intelligence for a few days, and I'm so impressed by how they made something even less useful and capable than Siri.
Every response is either 1) links to Google search, or 2) “I can't answer <question that's not what I asked, but uses similar words>”