the article headline immediately screams "financial gymnastics" to me so the rest followed from the quote.
4,022 karma · joined September 5, 2012
https://keerthik.github.io
cofounder bitgym, olin college alum, former digital nomad.
O1A visa -> EB1 Green card holder with only an undergrad degree.
Feel free to reach out if you aren't trying to sell me anything @keerthiko
the article headline immediately screams "financial gymnastics" to me so the rest followed from the quote.
Trying to incorporate it in existing codebases (esp when the end user is a support interaction or more away) is still folly, except for closely reviewed and/or non-business-logic modifications.
That said, it is quite impressive to set up a simple architecture, or just list the filenames, and tell some agents to go crazy to implement what you want the application to do. But once it crosses a certain complexity, I find you need to prompt closer and closer to the weeds to see real results. I imagine a non-technical prompter cannot proceed past a certain prototype fidelity threshold, let alone make meaningful contributions to a mature codebase via LLM without a human engineer to guide and review.
IMO regulation never was or is going to force this shift: it's already happening in unregulated ad markets, and is going to keep evolving in that direction because it's simply more effective/lucrative than ads done other ways.
> Break up Google. Don't tell content marketplaces how to run ads.
I'm all for breaking up megacorps, but there's no way a government like Vietnam can effectively accomplish that. The entire regulatory weight of the EU (90% of the non-US first-world consumer base) can't break up Google, so inflicting a series of wristslaps that hurt Google more than any small startup is the best way.
I'm no expert on the region, but I can't imagine a small video/social startup in Vietnam will be hurt more than Google by being forced to show a skip button after 5s on their ads — and generally speaking ads as a business model generally doesn't work all that well or mean much for small startups (<1M MAU), their survival and scalability hinges more on VC money and product-market fit than ad arbitrage.
There is another recovery option:
- increase the JPEG framerate every couple seconds until the bandwidth consumption approaches the H264 stream bandwidth estimate
- keep track latency changes. If the client reports a stable latency range, and it is acceptable (<1s latency, <200ms variance?) and bandwidth use has reached 95% of H264 estimate, re-activate the stream
Given that text/code is what is being viewed, lower res and adaptive streaming (HLS) are not really viable solutions since they become unreadable at lower res.
If remote screen sharing is a core feature of the service, I think this is a reasonable next step for the product.
That said, IMO at a higher level if you know what you're streaming is human-readable text, it's better to send application data pipes to the stream rather than encoding screenspace videos. That does however require building bespoke decoders and client viewing if real time collaboration network clients don't already exist for the tools (but SSH and RTC code editors exist)
What am I looking at?
When someone plays a game, the user's goal could be expected as "having fun for as much time as they want to." Being addictive is usually in service of that. A "slightly dark" pattern would be combining core addictive gameplay junctures with microtransactions (retry/next level/upgrade) — but in this economy this just feels like a basic mobile game business model. A moderately darker pattern would be making the game increasingly frustrating while still addictive, unless you perform a microtxn (eg: increasing difficulty exponentially, and charging money for more lives/retries or forcing more ads).
A "true dark pattern" would be sneaking things like push notification permissions, tracking permissions, recurring subscription agreements, etc. under an interface that looks similar to something the user doesn't read carefully and tries to get past out of habit, such as an interstitial ad with a "skip" button — but with a below-the-fold toggle button defaulted to "agree" and a "Confirm" button styled to look like the "skip" button at first glance.
> This groundbreaking new initiative will make child care available to all New Mexicans, regardless of income, by removing income eligibility requirements from the state’s child care assistance program
What you are referring to is the progress they made prior to this news.
More telling rather than any specific lines or phrases, to me, is that the tone gives off a strong "i don't have a single strong feeling about this topic, i'm just transcribing others' opinions and vaguely coming up with a C2A" vibe — no human who'd bother to write a personal-opinions blog would realistically channel that energy into an 850-word essay without LLMs doing all the drudgery.
a) convincing yourself its worth the effort: almost every time an adult runs into a confusing element of a new language, they find themselves calculating how many people in the world speak this language, probability they don't speak english and likelihood of running into this person and circumstance, and it's easy to justify giving up and moving on
b) avoiding forcing it into the framework of your first language: if you have one distinctly favored language already, it's very hard not to try shove the new language you are learning into the former's mold, and this can be counterproductive in learning most languages that don't share an ancestor with your favored one.
a) is greatly mitigated by forcing yourself to be in said context by living in a place prioritizing that language. b) is greatly mitigated by already being bilingual+ with languages from distinct origins (eg: mandarin chinese and english) before learning a new one, so you can place the new language on a spectrum with the ones you already know instead of confined by the rules of just one.
Scaled up slightly, we use it to build plenty of internal tooling in our video content production pipeline (syncing between encoding tools and a status dashboard for our non-technical content team).
Using it for anything more than boilerplate code, well-defined but tedious refactors, or quickly demonstrating how to use an unfamiliar API in production code, before a human takes a full pass at everything is something I'm going to be wary of for a long time.
If we were taking a walk and you asked me for an explanation for a mathematical concept I have not actually studied, I am fully capable of hazarding a casual guess based on the other topics I have studied within seconds. This is the default approach of an LLM, except with much greater breadth and recall of studied topics than I, as a human, have.
This would be very different than if we sat down at a library and I applied the various concepts and theorems I already knew to make inferences, built upon them, and then derived an understanding based on reasoning of the steps I took (often after backtracking from several reasoning dead ends) before providing the explanation.
If you ask an LLM to explain their reasoning, it's unclear whether it just guessed the explanation and reasoning too, or if that was actually the set of steps it took to get to the first answer they gave you. This is why LLMs are able to correct themselves after claiming strawberry has 2 rs, but when providing (guessing again) their explanations they make more "relevant" guesses.
Besides static puzzles (like a maze or jigsaw) I don't believe this analogy holds? A model working with prompt constraints that aren't evolving or being added over the course of "navigating" the generation of the model's output means it needs to process 0 new information that it didn't come up with itself — playing a game is different from other generation because it's primarily about reacting to input you didn't know the precise timing/spatial details of, but can learn that they come within a known set of higher order rules. Obviously the more finite/deterministic/predictably probabilistic the video game's solution space, the more it can be inferred from the initial state, aka reduce to the same type of problem as generating a video from a prompt), which is why models are still able to play video games. But as GP pointed out, transfer function negative in such cases — the overarching rules are not predictable enough across disparate genres.
> I think you could prompt veo3 to play any game for a few seconds
I'm curious what your threshold for what constitutes "play any game" is in this claim? If I wrote a script that maps button combinations to average pixel color of a portion of the screen buffer, by what metric(s) would veo3 be "playing" the game more or better than that script "for a few seconds"?
edit: removing knee-jerk reaction language
> "They threatened me with a minimum fine of $5,000 or five years in prison if I refused to provide the password to my phone."
this isn't real/legal/enforceable (as the law currently stands) is it? how does one protect ourselves against this turn of events upon entry when the immigration officer's claim fails the smell test?
I now live in NYC, and my local Home Depot rents out power tools for $20-30/day. The typical tool would pay for itself in 2 days at these rates, but it's still worth it to me as these 2 days are often 2-3 years apart and I'd rather not store these tools in my cramped 1BR.
What I would love is a community-run tool rental service where we can donate a tool + pay a nominal membership fee, and borrow tools for free. I am happy to donate a $120 circular saw/impact driver if I can rent 10x different tools once each for $5/mo for the next year. The closest thing I have now is my local hackerspace, which is great, but I often have to work on my projects at the space, which limits the kind of home DIY I can do.
i use it for very targeted operations where it saves me several roundtrips to code examples and documentation and stack overflow, not spamming it for every task i need to do, i spend about $1/day of focused feature development, and it feels like it saves me about 50% as many hours as i spend coding while using it.
Does geographic grouping data depend on the IP address? If so I suppose it would need to be extracted first before hashing the IP, and I wonder how much that weakens the anonymization.
it is absolutely possible that after experiencing the bigger pond, people can develop purpose in their "original" pond based on values like community and relationships, or even simply dislike the vibes in bigger ponds and want to undo as much as they can. this is a super valuable thing to society and humanity for the most part, as perhaps more change can happen this way than big things happening in big places.
personally i struggle with this, because whenever i re-enter a smaller ecosystem (including/such as the one i grew up around) i feel like everyone has a distorted view of the bigger pond and self-limit themselves, which is a contagious energy i can't stand.
the former group probably leads the healthiest & happiest life fulfillment while pursuing their interests — i'm heavily biased though because i too fall into this category and am proud of this trait.
the latter group consists of people who either spin their wheels real hard and more often than not burn out in their pursuit of being the best, or pivot hard into something else they think they can be the best at (often repeatedly every time they encounter stronger competition) like gates & co, or in rare cases succeed in being the best even in the more competitive environment.
this last .001% are probably people whose egos get so boosted from the positive reinforcement that they become "overcompetitive" and domineering like zuck or elon, and let their egos control their power and resources to suppress competition rather than compete "fairly" ever again.
i think there's a subset of people from both main groups that may move from one into the other based on life experiences, luck, influence of people close to them, maturity, therapy, or simply wanting something different from life after a certain point. i don't have a good model for whether this is most people, or a tiny percentage.
The "basic US presence" requirement of green cards has always been present in the validity clause alongside the 5-10year expiry date, and not committing immigration fraud and other basic requirements to maintain green card -- a comical number of European green card holders gloss over/forget this clause every year, that is made explicit to them upon receiving the card and proceed to forfeit their green cards by not entering the US for over a year -- that is not a revocation (implies a subjective decision made by an official), it is a lapse of validity (implies some pre-stated condition was fulfilled).