I’ve done this in pure Python for a long time. Single file prototype that can mostly function from the command line. The process helps me understand all the sub problems and how they relate to each other. Best example is when you realize behaviors X, Y, and Z have so much in common that it makes sense to have a single component that takes a parameter to specify which behavior to perform.
It’s possible that already practicing this is why I feel slightly “meh” compared to others regarding GenAI.
Next evolution: UX should allow end users to build the prompt, and then the tool is generated and deployed at a random URL. Tool also keeps original prompt, so any user can customize into a new (random URL) tool page.
If you start exercising in your 20s, and never stop, it will be so much easier to maintain fitness in 40s 50s etc. The challenge is that the benefits are not yet visible in your 20s (when you’ll probably be healthy and at a proper weight regardless). Gotta lay that foundation for older age though!
EDIT - I misread the comment. It’s never too late to start, just be careful for injuries as that will block your ability to exercise.
As fascinating as these tools can be - are we (the industry) once again finding something other than our “customer” to focus our brains on (see Paul Graham’s “Top idea in your mind” essay)?
LLMs are good at creating single use documents, like a pitch deck used for one prospective customer (never to be used again). But for long lived documents, on which future work builds atop, the bar is higher and the value of LLMs is more grey.
If the “C” is broke, it seems like there won’t by any rich people. In other words, if the masses are poor and jobless, who is sending money to the rich?
Have you tried non-running (indoor) cardio as a regular replacement? Jumping jacks, squats, jump squats, mountain climbers, even jogging in place. I do cardio everyday but running is only every 4th day.
Of course, if you have access to equipment there’s always stationary bike and other machines like an elliptical.
I see 100x used quite a bit related to LLM productivity. It seems extreme because it implies one could generate a year’s worth of value in a few days. I would think delivering features involves too much non coding work for this to be possible.
Societal norms are the problem. We’ve normalized unhealthy food. Big business has figured hit that if they make processed/refined foods easily accessible, you will buy them. Nobody wins here except shareholders, and only those shareholders in good health who don’t have to compete to get access from doctors who are overbooked with patients who have manifested a chronic illness that is statistically correlated with the aforementioned food.
You’ll also want to familiarize with “git apply -3 <file name>”, for when a diff can’t be applied cleanly. It will try “harder” to merge (three way method) and if it still fails it invokes the conflict merge “UX”:
Once I discovered how git apply can take diff files (or patch files) as input, I stopped using git stash in favor of plain old files. Easier to list and browse the contents of prior edits, also you can grep the files as method of search. I’ve even found myself copying and editing the diffs before applying.
I’m not sure it’s possible to have the full reasoning in your head without authoring the code yourself - or, spending a comparable amount of effort to mentally rewrite it.
Interesting that if people took their health more seriously, the food companies mentioned here would struggle to survive regardless of pricing power. Total addressable market would plummet. I suppose one way to accept the status quo is that these companies are self levying a vice tax.
This is a very interesting abstraction (and service). I can’t help but feature creep and ask for something like Athena, which runs PrestoDB (map reduce) over S3 files. It could be superior in theory because anyone using that pattern must shoehorn their data stream (almost everything is really a stream) into an S3 file system. Fragmentation and file packing become requirements that degrade transactional qualities.
I’d wager that the algorithm is slightly eager to throw a consistency error if it’s unable to verify across partitions. Since the caller is naturally ready for this error, it’s likely not a problem. So in short it’s the P :)