It's been a while since I've set up a windows machine and this may already be mentioned, but when I sign up for signal I got lots of warnings that were like "warning if you lose your phone and encryption key you will lose your data"
That way I know what I'm signing up for.
Just put "encrypt? Yes no" in the on-boarding flow and let people know what the risks are and what they may be protecting against. I'd probably default to off because people don't read wizards and the last thing someone wants is to lose their entire HDD because they accidentally made a decision they didn't understand.
And maybe for a certain period of time they can nudge users to read about encryption and decide if it's right for them, or just easily disable that nudge. Maybe even basic education like "if you find yourself forgetting your password often then maybe encryption is not for you" or something like that.
Windows is already optimized for extracting as much value from customers as possible, may as well help them make at least one informed decision.
I'm interested in seeing how this changes folks' workflows.
For me, at work I use opus to plan, brainstorm, grill, ask questions about my codebase, etc. It is pretty good about understanding the codebase holistically and providing architecturally clean solutions that actually work. Then I use sonnet as a plan executor and it does well. Follows instructions and runs tests and just overall does great.
At home I make some toy projects using opencode go (I've standardized on deepseek 4 pro as my opus replacement) but it's pretty obvious from the amount of times I've had to fix or revert a change that broke something that it's no opus. I got similar results with kimi. Have not played too much with Qwen.
So I'm wondering what I'd use to get a similar stack at work. Folks say that this version of glm is basically Jan 2026 opus pre me f. Big if true. So would I use GLM for plan and Deepseek v4 pro/flash for execution? Or maybe Kimi or Qwen? I know I'll probably never get as good quality code as I do at work but I'm just toying around here.
I've found the prompting needs are drastically different from the latest frontier models to the latest open weight models. I can be much more vague and talk about an end goal with the frontier models vs needing to be more prescriptive + have a workflow on the open weight models. This gap continues to close, but the level of abstraction I'm working on with the latest models continues to move much higher.
I use glm for all code investigations and top level system design of all kinds, and then present finding to confirm and act upon to opus. everything that burns token goes there.
the finding aren't always accurate, but it saves ton of opus token
likewise I have google ai from my photo storage, so I give claude / opencode a skill that uses gemini (agy now) command line for web searches, using their flash model line.
This is interesting because you can't prove you prevented something from happening. It never happened after all. Hank Green has an interesting video on preventative measures are often unsung, because how do you know?
Kids in Africa who would have otherwise died from measles, but didn't get measles in the first place because some people decided to fund and make vaccines and thousands of unnamed and unrecognized volunteers helped distribute and administer these vaccines...
These kids are alive and they don't even know that they wouldn't be alive if it weren't for some people they may never meet, and whose names they'll never know.
Also tangentially, reminds me of an excerpt of a book called The Trigger Effect I read as a kid, which said something along the lines of, when you use your car's brakes to prevent an accident, your life depends on whoever assembled the brakes and quality controlled them. You'll never know their name, and they'll never know that they saved your life.
Learning doesn't happen in a vacuum. Even pre-LLM days where I'd scour stack overflow for the solution to one problem, I'd inadvertently learn other random stuff while looking.
Easy to say, but every bank I've had the (dis)pleasure of doing business with only ever issued a Visa or Mastercard so it's not really feasible to just "stop using them"
I think it's kind of counterproductive to call out "this comment sounds like AI."
If you're right and it is AI, the comment only points out what many of us can already see. If you're wrong and it isn't AI, then the commenter you're responding to can either argue with you about how it's not AI or just say nothing because there's nothing of substance to say.
reply