David E. Weekly; ran R&D teams for Facebook, Google, and Capital One, founder of Medcorder, Drone.VC, PBwiki, Mexican.VC, Hacker Dojo, SuperHappyDevHouse. Long history of hackery and tomfoolery, MP3s, computer graphics, flying helicopters and planes.
A friend of mine got phished and I did a teardown of the infrastructure components that were used to execute the attack, that executes a Browser-In-The-Middle attack with an encrypted websocket. I've already reached out to the involved infrastructure providers, but haven't yet heard back from all of them yet.
On the play button, I found your issue; I'm guessing you had Silent Mode enabled and I hadn't set the audio session to 'playback' from ambient. (You would have heard the sweep if you turned off Silent Mode.) I've now set playback mode and I have verified it now plays back just fine even in Silent Mode. You can verify it on cyrinx.org. Thank you for the bug report!
On licenses, what's your recommendation for best practice for an LLM co-authored project? Apache 2.0 or MIT seemed the best fit.
UPDATE: On the play button, I found your issue; I had tested on Chrome desktop and it worked fine, but I'm guessing you had Silent Mode enabled and I hadn't set the audio session to 'playback' from ambient. (You would have heard the sweep if you turned off Silent Mode.) I've now set playback mode and I have verified it now plays back just fine even in Silent Mode. You can verify it on cyrinx.org. Thank you for the bug report!
For human in the loop to be effective, the human needs to actually be performing some substantive action, giving real guidance and critique and pushback. If the human only ever accepts the default plans then not only is there no understanding but the agent should learn to stop asking. It is not learning anything from the human, after all.
One thing that I look at is pushback rate: what percentage of the agent's proposals are rejected or critiqued? If it's below 5% I have found I have gotten too credulous and I am no longer closely following. Danger! If it's above 50%, I have clearly not given the agents sufficient context to perform the task and need to update my harness and instructions.
Who watches the watchers? I can imagine a guard dog process that halts the session to yell at the human if it detects complacency: if the human is providing too few tokens per minute of new context relevant to the task.
That does seem to be the path Apple is following here. Have a local model that can answer most things and then have a fallback of cloud options when they request is too complex. The cleverness of this strategy has been overshadowed by the incredibly poor quality of their local models. It will be extremely interesting to see what next month holds and whether Google helped fine tune an Apple specific Gemini / Gemma model for their devices. Bonus points, of course, if they unveil the M5 Ultra Studio with half a terabyte of RAM to be a local "cloud model" (the true fantasy here of course would be Apple building something a little like openclaw where from your phone you could give commands to your Home Apple server). They could probably get away with charging $20k for it if it has sufficient tok/sec. If that happens and is successful one could imagine a straight line path in the next two generations to bringing the cost and form factor down to the point where some of the form factor of an Apple TV becomes everybody's home inference server / agentic HQ. Sovereign AI for everyone!
And by "places orders" we mean "helps TSMC acquire plots of land on which their next facilities will be constructed" kind of level of scope, timing, and commitment.
Note that this is how the FAA has been doing its written tests for years: you go to a proctor test facility and through a metal detector. The entire thing is videotaped.
The only path to cool is through cringe. You can do it!
I think a lot of people after a random compliment might be wondering if the other person is trying to get something out of them, like a date or a business deal. When it becomes clear that you're leaving it there with no expectation of reciprocation they can truly internalize the compliment.
I think a reasonable proxy for "care" is genuine curiosity about who a person is, how they came to be where they are, what makes them tick, coupled with a general desire for positive things to happen to them in their life.
If you believe that most people have things about them that are fundamentally interesting, you will put effort in to find those things, and you'll generally be successful in finding them. If you have a belief that most people are fundamentally uninteresting, you will not put in the work, and your beliefs will be validated because you will fail discover anything of interest in most people around you.
The act of earnestly wondering about a person is very flattering - this will help people open up to you and share their interestingness. If you can layer that with an additional thought of what useful thing you might be able to do for them, such as another person who might be able to help them, or a piece of advice you can give that might save them time, you will develop a reputation as a person one should talk to.
This then will increase your hit rate on being useful which is a nicely positively compounding cycle.
My read on the book was "humans are really good at telling if you genuinely care about them or not and will respond well to that, so you should genuinely care about the people around you, and good things will result from that overall, especially if you're not super mercenary about it."
Bill & Ted said it most pithily: be excellent to each other.
The pricing models that are published on AWS' website almost certainly have almost nothing to do with the pricing models that are discussed behind closed doors for a $100 billion commitment.
Having just checked my child in for their doctor appointment, 90% of web software would be dramatically improved by using very boring best practices and readable and accessible web practices.
GPT 5.4 is the surly physics PhD post-doc who slowly and angrily sits in a basement to write brilliant, undocumented, uncommented code that encapsulates a breakthrough algorithm.
Opus 4.6 is the L5 new hire SWE keen to prove their chops and quickly turn out totally reasonable code with putatively defensible reasons for doing it that way (that are sometimes tragically wrong) and then catch an after-work yoga class with you.
Socials: - github.com/dweekly - linkedin.com/in/dweekly - x.com/dweekly