a-dub

the dynamics of discovery on a population level are also not fully understood. there's no doubt in my mind that the machines will grab all the low hanging fruit in terms of novel ideas through mechanical recombination of existing public science, and they will do so faster than humans can. what is not understood is whether they will plateau.

today they benefit from human willingness to share publicly and improve ai systems, but once those systems start to threaten their livelihoods, will those dynamics change? can the machines push the frontiers or is the global network of human creativity and tenacity necessary?

i like to write and then workshop with the ai. i find that i have to push back on tendencies it (has less of these days) has to rewrite my work in massive ways. it also can help with writers block sometimes, or parts of a paragraph i'm not that interested in.

> There’s a wonderful book, How to Prove It by Daniel Velleman

the name sounds familiar but i don't think i have read that one, i did enjoy "introduction to mathematical reasoning" by eccles.

personally my relationship with mathematical proofs has been complicated. it took some work to understand basic proofs (dedekind cuts, ideas vs. instructions with mathematical notation), but all of the theory of computation proofs, which supposedly are difficult for many, were completely intuitively easy for me.

i think mathematicians are facing a similar confusion as computer programmers. the medium used to require precise thinking and the simple act of reading, writing and composing it was a mechanism for thinking and learning. in the llm era, the question is: should there be a new mechanism and if so, what should it look like?

i bet if you mmap'd a big file and then dirtied a ton of pages and then exited and watched the file on disk, you'd see writes occurring after the process exits as well.

yeah in that case, use a binary for "did you do it that day?" and a measure for "how much of it did you do that day?". loop habit tracker supports both, i bet this one does too. a built-in stopwatch to make time tracking specifically easy would be cool.

i have done things like "do x at least once three times a week" and then had a separate counter for how many times. it's good to track intensity or progression, but i think the most value comes from "did you do it at all that day?" and then sticking to a plan that involves n days a week.

the finer grained tracking can contribute to "i am not on target, so maybe it doesn't matter" sort of thinking where maybe the most valuable thing is "i am setting time aside n days a week and sticking to at least attempting the activity on that schedule."

ymmv of course, this stuff is deeply personal!

i don't think that computer vision gates or even an ai agent trying to enforce accountability are going to help all that much in that case. maybe try a therapist or a real-life accountability partner.

i would say, be generous with yourself in terms of tracking showing up at all since that alone is huge (and most important for sustaining the behavior). if you notice that you're slipping and showing up late too often, maybe create another habit for being on time?

i don't understand the rationale behind cheating one's own self improvement journal. what's the end game?

writing well still works. it also still can be quite hard depending on the circumstance.

> You watch input to output delay on a scope and look for outliers.

better to acquire these or use timestamped gpio and compute real statistics- but for the sake of illustrative metaphor, sure.

this is exactly what these weird operating system age tracking laws are codifying which is terrible. it forces consumer operating systems to become trusted computing bases that don't work for the consumer.

upgrading to adult content when a credential is provided at the protocol level rather than forcing the operating system to track the state means that consumer operating systems don't have to be secured against their own users.

i wonder what they're all up to. i imagine some are scraping datasets for pre-training, others are probably real-time scrapers looking for security bugs, even more still are agents working on coding tasks and looking at the kernel. also interesting to think about solutions: does everything need to be optimized now for weird access patterns that proliferated ai creates? do the ais need to have behavior trained in to be better netizens? is this the end of anonymous browsing and the beginning of an era where one has to attach an identity to all requests? or the end of community hosted free information services more broadly?

i still think this should be solved at the protocol level. if you want to access the adult stuff, you have to staple a non-identifying zk-proof-of-age to your http/spdy/etc requests, otherwise it all just operates as usual.

san francisco had that cool, edgy, gonzo art scene happening in the 90s. also survival research laboratories and their famous pitching machine, the omnicircus, chicken john, laughing squid, critical mass/commute clot, santarchy/santa con (before it became the weird frat thing it is today). similar things in brooklyn in the '00s like the madagascar institute, black label bike club, idiotarod, etc.

in the 2000s, there was kind of an interesting tech meets countercultural art thing where a lot of the more interesting tech people would pour their disposable income into interesting burning man art projects (nimby's ddi, jabba's sail barge) but it does seem like in recent years with the real sf affordability crisis that came with internet business actually taking off the creativity went down and the money seeking people showed up. i guess that's also true of sf (and brooklyn) more generally.

it was a fun time. i was a swe intern in the mid 90s and would marvel when the engineering team would disappear to the desert for a week and come back with colored hair. my boss drove a psychedelic school bus to work sometimes. in south of market on 6th street in sf, there was a building that had furniture welded onto the exterior like it had all come to life and was trying to escape the building.

the internet is home to plenty of predatory cesspits, but the real stuff is still there too. it's not as convenient or addicting as content that doubles as a funnel for the ad machines, but the real spaces and media that aren't dominated by the advertising industrial complex still exist- just like in other forms of media that preceded the internet.

last time i built a crud app (oh maybe about 24 years ago) i made it a point to do this. i remember watching the payroll people fly through their terminal mode vax vms applications with such speed and dexterity that it would make any unix admin well versed in the art of the command line blush and it made me think "oh yeah, all these 90s point and click guis got it all wrong. if people are required to make heavy use of a system at work, they would prefer a learning curve followed by speed, comfort and dexterity over an easier learning curve that trades dexterity for discoverability."

it was a webapp framework, but... all browsing/listing screens included row ids and a focused text box- so that typing the row id and enter would select. all action buttons had an underline to signify which ctrl-shift hotkey triggered them. all edit screens defaulted focus to the first editable textbox and at no time was the mouse actually necessary. finally, load times were optimized to target 75ms.

amusingly, the user feedback was "the keyboard control is pretty good but can you please make it faster."

what i thought was chrome ended up being critical to the users not being miserable.

i think ai is going to turn cybsersecurity into a real-time affair that looks a lot more like high frequency trading.

ahh. kde has all that built directly into the screenshot utility. highlighting, text, cropping, etc.

it is kind of interesting when you think about it. what utility does something like paint actually have? did it ever have any real utility? (icon editing?) it always kinda felt like a demo that came with early versions of windows that didn't actually do anything useful.

hanlon's razor comes to mind. which trees are these? weird device trees that have complicated third party licensing nonsense attached?

i remember jumping through crazy hoops to interact with a google open source project years ago. i wouldn't be surprised if it's just megacorp bureaucracy.

cast it to an integer and check for equality?

requiring bit identical goes beyond what floats promise though.

> What is interesting about it?

did western society engineer a junkfood-industrial-complex so addicting and so en-sickening that it ultimately required a sort of junk food methadone to wean itself off?

that is the interesting question. also the effects of weight loss vs. the effects of disengagement with the engineered to be addictive "food products" side of the western diet.

ahh. thanks! i read localization, thought of 5g beam forming and thought "oh this must be wifi sensing 5g beam forming edition." i stand corrected, thank you!

uwb is pretty neat. in the last few months i started getting 3/4ths of a gigabit of downstream bandwidth on a lot of nyc streets on my ordinary verizon plan.

existence or absence of drama around getting the right equipment is usually a very strong indicator of technical leadership quality.

ahh yes, the early 90s. computer shopper, datanet, pc magazine, a $59 seller's permit from the state of california that came with a magic number that would make distributors sell to you at cost for both computer and bike parts, automated touch tone fax back information systems, the absolute cheapest xt clone that could be assembled from parts from the computer shopper, ups cash on delivery, genie, prodigy, the bbs scene, the dreaded "rma number" and the subsequent wait, searching the space of physical card positions, jumper settings and tsr drivers that would allow the mouse, the soundcard and the game to all work at the same time, the borland compiler suite, mark williams coherent, os/2 2.0, trumpet winsock, slackware linux floppies...

not a bad time to be a kid.

i'm curious what this gains in practice and usability over standard chatbot slack integration.

it is definitely the claudiest. it's weird, it's like there's a spot developing in my head next to all the other spots where i park mental models for how people write, but instead of being for a person, it's for the terse-staccato-prosodic-diarrhea that claude generates in readmes by default.

it's just like a phone. don't buy a crappy one with firmware of unknown provenance. make sure the one you do buy has an active and effective effort that you trust that ships timely security fixes.

hm. in linux you have MAP_POPULATE which forces a prefetch where macos relies on page faults and lazy loading. if MADV_WILLNEED doesn't help, maybe readv to vector read directly or mmap+writev (write to a dummy fd, with an iovec for each page, hopefully resulting in a one-syscall-big-pagefault for mmap.) maybe also experiment with a loop that just reads one byte from each relevant page after mmap but before real computation?