717 karma · joined August 18, 2015
https://github.com/re-cinq/claudit
What's amusing is that I specifically made a comedy license for it, because I thought trying to protect the IP of such a thing is madness in the AI-native era. Then within days I find out it's a massively funded startup!
https://github.com/re-cinq/ai-native-application-license/
I've got Claude Code adding Gemini and OpenCode support to my tool currently.
It'd be interesting to see the experience of folks in continental Europe, where I gather wine with meals is much more commonplace.
We run Concourse London User Group, and the first few instances were like some kind of Jenkins survivor's support group.
Long trips in a Tesla are totally doable.
Hear, hear!
If you can accurately estimate how long a software task will take, you should have already made a reusable component or automation to generate that code.
Estimates are meaningless. I've seen PhDs waste endless hours faffing with estimate-calculating spreadsheets.
Fundamentally, the universe is unpredictable. Chaotic and complex systems require their starting conditions to be measured to an infinite level of precision to be predictable. The Heisenberg principle means this is, as far as physics can tell, impossible.
On a more practical macro level, a complex adaptive system becomes unpredictable once three feedback loops are present (the three body problem is related).
Modern computer systems are unpredictable because we cannot predict the interplay between levels of abstraction.
It does not matter how smart you are. Unless you have perfect knowledge of all levels of abstraction in the system below you, even in a macroscopic sense you cannot predict the future. Precise estimates will be inaccurate.
Confidence ranges and superprobabilities are of some use. Discrete and precise estimates are an utter waste of time at best, and are dangerous and misleading at worst.
EDIT
Written on a mobile waiting for plane take-off, hence lack of citations. For more on complex adaptive systems, I recommend the works of Murray Gell-Mann and the research output of the Santa Fe Institute, particularly Scott E. Page.
Don't do sprints. Have a continuous backlog. Don't do overtime. Don't make estimates. Always do the simplest thing. Only ever do the most important thing, as defined by the stakeholder.
I've written and talked about this at great length. The fact people suggest agile gives you burnout reinforces my experience that Scrum is largely misinterpreted and people incorrectly focus on sprint commitments. If Scrum is so commonly misinterpreted, it is flawed.
https://www.linkedin.com/pulse/scrum-makes-you-dumb-daniel-j...
I don't believe that they're likely to pick up something driving straight at them, as it's not something the neural net has been trained for.
My Model X often alerts me to parked cars, thinking I might drive into them :P
TLDR he makes the argument that entropy is a measure of the probability of probability distributions.
Imagine a customer has £100 in their account. System partitions. Customer withdraws £70, hitting one DC. Customer then hits the second DC, this time withdrawing £50. Each DC thinks the transaction is valid, and so serves it.
Later when the partition is restored, events are played back, and divergent history is detected via the vector clocks - the two withdrawals are not causally related. Remediative action can then be taken.
Transactions prevent bad things happening, but require CP semantics. Eventual consistency allows AP, allows bad things to happen, meaning you have to be able to detect them and clean them up later.
---
I've not really used Kafka, so couldn't comment on that. I did some work for a customer that involved multi-DC microservices with isolated databases (ie one DB per DC). We used event sourcing with vector clocks to do manual reconciliation of the databases including during partition. Reconciliation involved custom logic depending on the event type, so not sure how a transport mechanism like Kafka would handle that.
A talk about the approach is here: https://youtu.be/hYVh8PbbeJw
I get to vector clocks at the end.
Let's not suggest that Autopilot is safe completely unsupervised, and let's also not suggest that all drivers always look at the road 100% of every drive they take. There is a realistic middle ground, and I'm very grateful to have Autopilot to be an extra set of eyes.
Turning on windscreen heater, heating, and all that stuff is pretty straightforward, the on-screen buttons are always in the same place and are next to the bezel so you can anchor your hand. Plus cabin heating can be controlled via clickwheels on the steering wheel.
I think maybe the seat heaters are a little distracting. You can only tell their state by looking at the screen, and they cycle through settings. That's something I'd like a button for, or at least to be able to control via the clickwheels.
Edit
Studied HCI and Usability Engineering at university, did my dissertation in usability of web systems, and generally share your concerns. Not aiming to dismiss them, more say that in my experience it's not as bad as one might expect.
Reading Judea Pearl's The Book Of Why certainly sobered my outlook on AI.
It's a bit like networks bragging that they have 5G base stations before any 5G phones were created.
I gather that European Model 3s will have CCS sockets (a superset of pins of Type 2 Mennekes) so they will be able to charge at other networks' stations.
One-mile-per-minute definitely sounds like something was wrong. Did you try other charging stations? How full was your battery at that point? If you've seen that with multiple stations when you battery was neither nearly empty or nearly full, I'd suggest getting the car serviced as that's not normal behaviour.