authorfly

What changes did you make?

LAMP with dynamic webpages (I assume your approach) works just like it ever did (besides SSL)

But are you really keen to make a PHP dynamic webpage application where each page imports some database function/credentials and uses them to render html?

Can you keep the behavior of fluent userflow (e.g. menu not rerendering) that way? Only with minimal design.

When in 2006 most webpages had an iframe from the main content, an iframe for the menu, and maybe an iframe for some other element (e.g. a chat window), it was fine to refresh one of those or have a link in one load another dynamic page. Today that is not seen to be very attractive and to common people (consumers and businesses), unattractive means low-trust which means less income. Just my experience, unfortunately. I also loved that era in hindsight, even though the bugs were frustrating, let alone the JS binding and undefined errors if you added that...

Yeah, you are very right.

If you start peaking success, you realize that while your happy path may work for 70% of real cases, it's not really optimal to convert for most of them. Sentry helps a lot, you see session replay, you get excited.

You realize you can A/B test... but you need a tool for that...

Problem: Things like Openreplay will just crash and not restart themselves, with multiple container setups, some random part going down will just stop your session collection, without you noticing.. try to debug that? Goodluck, it'll take at least half a day. And often, you restore functionality, only to have another random error take it down a couple of months later, or you realize, the default configuration is only to keep 500mb of logs/recordings (what), etc, etc...

You realize you are saving $40/month for a very big hassle and worse, it may not work when you need it. You go back to sentry etc..

Does Canine change that?

I would say I wouldn't agree, but maybe our language in common is why we don't agree - for example, the idea that coffee is 80% takeaway... this global report puts it at 65% in cafe. My experience in most of europe is about 75% in cafe: https://www.grandviewresearch.com/industry-analysis/cafe-mar...

So maybe your use of coffee is just different to you know, the old world ways that developed it.

I totally agree with your suggestion for ordering, self-reporting the success and the "if x then imagine y!" you discuss and I get you didn't ask for criticism, but in your last example - the value of coffee to you might not be what it costs to make. The value can be a lot more for some people than others:

- Utility-wise: If you need to perform at work.. even if having it every day means the coffee nolonger changes from your original no-coffee baseline..

- Social: My girlfriend doesn't really go to the expensive coffee shop because it's worth the 2x marker. It's because she can chat about relationships, have security in her relationships with her friends and they can designate this as an unscheduled-long time to talk through issues

- Regulating your emotions For me at least buying something small and pleasant can avoid troughs/peaks. In fact overstressing the financial ratio of cost:price of a product causes these troughs for me. That's also why poorer people often buy small but tasty things, even when it's pricy for them, it regulates their mood up enough to slog through the rest of the 1am shifts with difficulty seeing the future (I have been there...)

That's why I'd be wary of having an app pressure you not to buy something. You don't always know the real reasons your body gets you to do things.

As you age your interests and curiosity change, in ways you often don't see until later.

Very common in computer science contexts. Young undergraduates always pick up the new tech and make something that seems alien and wrong first. It's not even the masters students.

Possibly the same Kiro - Agentic IDE post would have been as interesting to you as the launch of Atom or something related to VS Code, etc.

Would you say LessWrong posts are dogmatic?

I think you're right. About 20-30% of product hunt products were AI in late 2022 for example, but very few of them then used .ai.

your perspectives

90% gravity / 90% done is a bad perspective. You don't really have data on how long the false positives would have taken to succeed.

Unfortunately, getting from 0 to $100 in revenue isn't always much harder than $100 to $200. Even though the first $100 is insanely hard, so can the next $100 be. People don't like to talk about this in the startup space but the statistics at innovation centres for startups etc are insanely clear: at any given revenue, most startups are only going to keep it above the current value for part of the next 6 months, then crash and burn. And upwards line, let alone exponential growth, is an outlier (and still leads to revenue decline the next year most of the times it happens). Unfortunately, finding something that works (a fly well or net ROI positive channel) often takes so much time maintaining you take your eye and progress off other aspects you would have focused on for success and it becomes a revenue wave when it stops working rather than a stable way to grow.

Second problem is young people think they won't tire.. after three years almost anyone has a period of burnout. Atleast althetes like Olympians or Baseketball players have off seasons, with startups, it's not so. This is why having great people with you who flux at different times and take care of stuff you take your attention of increases success, if you can keep conflict minimal. Yet, most young people also don't want to relinquish control, if they are honest.

That's my view.

100%.

Dealing with 5 year old code is okay now. In 10 years it will cause you pain. It's like going back to an ex after you lose attraction. It never works out and hurts both parties (in this case your users, because you won't maintain it or want to pretty soon)

Good advice and beware the difference between docker export (which will fail if you lack enough storage, since it saves volumes) and docker save. Running the wrong command might knock out your only running docker server into an unrecoverable state...

You'd like Milton Friedman and his take on pilot Unions too.

What's your patient/flexible technique? Let me know.

And you are not a liar - but your claim isn't true at all in Europe - see increased per-flight legislated fees and the loss of budget airlines. Price of flights between 2 destinations has increased by 25-40% in the last 5 years in most of Europe.

Thanks to efforts like increasing the per-flight fees "because of high inflation" (these fee increases are still going up several years later): https://www.nationalgeographic.com/travel/article/why-travel... and loss of airlines like Flybe.

You can still get the madly cheap hop flights, but they are often pricing in income from our flights (or accepting even negative returns by pricing above the per-flight fees) because the planes need to be where you are going to fly a profitable flight later.

So the old status quo where genuinely cheap flights could be booked on a 7 day basis (e.g. cheap thursday-thursday flights) has been replaced by convoluted patterns to get the cheap flights (you usually need to leave on Monday, return on Saturday (if your source airport has lower demand than the destination) or vice versa. I suggest based on flights in the US becoming cheaper, that this is due to government intervention.

I get saving the environment and all that. But let me pay more taxes monthly, don't charge the airline £15 minimum making a bunch of flights unviable. Don't make booking a holiday or conference flight so unpleasant and annoying. I always have to tradeoff wasting a day or two with paying 50-150 euros extra.

It's not the worlds biggest problem, but making that decision is a regular additional dilemma I didn't want in my life. I wish for the days when you could get just normal timetabled flights at good costs if the month (e.g. February) was unpopular for travel. Now those months really aren't cheaper.

Single also differs from being in a childless couple e.g. in your early 20s.

To get on the ladder today, 5 years sharing rent is priceless. Then once you do, you get child benefits. Many people are single late in life too. So I don't think it's something you can equate.

What packs do you buy where it is much more cheaper?

Frozen > 5kg or so?

This is part of why France lost status for large events and organisations in the early-mid 20th Century until now. Quite large political swings at regular intervals and fairly ready-to-protest population = not ideal for basing stability. And it doesn't matter that France was far more stable than most countries, it was simply much less stable than America, the UK and regrettably, Germany. Frances solution seems to be not to really compete for international industry or science, but to focus on French culture and language to fit their tourism strategy.

So that adaptive system sounds really cool and useful for those bursts of formative assessment. And if the baseline is 100% boring textbook reading, ofcourse that is better. But if they work 9-5 every day, why don't schools do that at the moment do you think?

I don't mean to be negative, but can you name a research paper where a normal students 7 hour school day was changed to formative assessment all week long for a semester or more, and the grades improved? Or identify why students on Duolingo don't tend to learn as well as those taking in person classes? Why does the effect of the Testing Effect/formative assessment tail off when it is used more and over greater time periods?

The point is, the research on learning methods usually researches them in isolation and in 5-15-60 minute study sessions. Maybe doing 15 minutes of formative assessment improves your grades, but puts you off doing other work or makes you less effective in later lessons, when done regularly. There is a lack of useful longitudinal research in Education for the depressing reason that the two main effects are your genetic background, and private teacher attention.

I don't think kids are not motivated about math because they are bad at it. It doesn't seem like a subject that most kids see themselves using every day, because the adults they see around them don't seem to (and in fact largely don't) use Maths. While I value and enjoy Maths, and know how knowing maths makes a massive difference to your experience or life, kids can't, and really won't have good reasons to understand or be motivated through that. e.g. it's not a fear of negativity (though in class under pressure it is). It's an absence of motivation in the first place.

Research shows low-stakes assessment is when the moment of learning often happens

Okay, but psychology research like this needs to be taken with context on long term behavior. Is this advice practical? No, not in this case. The problem is that very few people can actively learn that way for multiple hours in one day, let alone be motivated for studying in that intense way for subjects they do not like. Any barrier will stop them. That research applies to cramming for exams, when stressed, which most EdTech isn't really useful for (at that point, people are learning intensely anyway). This human nature has not changed.

And your suggestions (not to be rude) are to excite kids about the subjects they find least motivating. But the problem is already motivation.

Youtube is motivating because it is fun. It might not work, but it fits the motivation level you have.

Why not forget about the strong moats/tendencies of EdTech projects to fail to revolutionise, forget them targetting parents, and either pick a subgroup of learners who actively do want to change how they learn, or work on the motivation problem? Why are some kids so much less motivated to learn Math than others? Should we change that and how can we?##

Here's an example I find interesting: The western world accomplished bringing Biology from dominantly male to an equally balanced subject. How did that change happen? We benefit from all these women researching in Biology. Could it happen to other sciences? Or the reverse? These kind of questions interest me.

Can I engage you on this as someone who once shared your view? Not to say I believe my view is better now, but maybe you can learn from my experiences.

Not everyone has this reaction, because what they have been exposed to shapes how that content will affect them.

Specifically people who have been victims of serious assault or even witnessed that can have a much worse, and irreversible reaction to you when seeing things that make those memories come to the fore as recurrent, intrusive thoughts, which then affect their behavior and lives. That is really what the restriction of content should be about if anything: helping people avoid things they want to avoid.

The people who have struggled (especially at a young age) with real trauma often come across as distant, quiet or anti-social; sometimes they never were so before. But often, our community where this behavior is more normalized, is where those people come, even if they don't have a primary interest in the community, to feel normal again, while still feeling fearful or full of empathy. You may have trauma, or not, depending on what violence you faced. However, even with violence, people react in wildly different ways, for one, women are much more anxious and cautious after feeling at risk or violated than men, so you really cannot assume that how you feel represents how a woman would (for evolutionary sensible reasons). Meanwhile, men often suppress their emotions (at a truly deep level, killing their relationships).

The problem with saying that prohibition necessarily means they will encounter the material in an unsafe environment is that, someone who has been assaulted or abused is already in an unsafe environment, everywhere, in their mind, and for legitimate and rationale reasons. The world is different when you know police will generally not deeply investigate a serious crime, when one has been personally been conducted against you. Seeing content like that, can prolong or make permanent that state of being, which can leave to bad and convoluted consequences later on. It is easier to understand this if you have children or have seen real pain and suffering with someone you love too, that can give you the empathy to understand this reaction.

It is hard to understand psychological damage unless you or someone you truly love and have strong empathy with goes through it. Until then, it's hard to understand or imagine at all how other people might be affected by some things. They will not always have your reaction to content which is extreme. I do not agree with prohibition, but do consider that others can have different reactions to you, ones you possibly cannot imagine.

Put another way, many times, we label content extreme not because it is extreme for everyone. We label it, because for some group of people, at some point, it could set their own lives back a lot to encounter it, and these people are already suffering more than the average person. It's about helping them avoid more pain.

Obviously this does not apply to all content, but for your examples, it does. Do not imagine there are not blue collar workers who have seen close friends suffer similar pain to the fate you mention, haunted by it. Men who would break at the knees at the sight of that kind of video. There are. You brush shoulders with them on the street. We can understand the dark sides of humanity through history and the written word (which I believe should be fully unrestricted), but not everything needs the very human, memory-provoking visual element.

the release and research cycles are still contracting

Not necessarily progress or benchmarks that as a broader picture you would look at (MMLU etc)

GPT-3 was an amazing step up from GPT-2, something scientists in the field really thought was 10-15 years out at least done in 2, instruct/RHLF for GPTs was a similar massive splash, making the second half of 2021 equally amazing.

However nothing since has really been that left field or unpredictable from then, and it's been almost 3 years since RHLF hit the field. We knew good image understanding as input, longer context, and improved prompting would improve results. The releases are common, but the progress feels like it has stalled for me.

What really has changed since Davinci-instruct or ChatGPT to you? When making an AI-using product, do you construct it differently? Are agents presently more than APIs talking to databases with private fields?

Artistic rendering, yes. It's not usually something new, afterall. Incredibly divulged from really rendering what scientists want to communicate about something somewhat visual with some basis on reality (and real images) while using size, shape, contrast and colors (sometimes with artistic license) to communicate effectively a scientific idea.

privately-owned surveillance cameras that are "voluntarily" offered to law enforcement

You may find interesting to know that in the case of crimes being committed near or next to a business, especially independent businesses, their priority is often to minimalize risk to them and their business, and so they do not provide any recording, or use rolling footage which wipes every x days (often just one day). That way the footage is useful to them, but they do not have any additional obligations.

That's a real difference when something is your final responsibility too (as team lead or an architect). You think of it differently, you predict and anticipate changes better. It's like taking care of your kid vs your kids friend.

People matter, you need to be communicating to keep communicating up, not get stuck repeating yourself and losing your soul. It keeps accountability up.

You need diversity in your team, even though that has costs and friction. You need experienced older people, not because of some advantage (there are some, but if young, you probably won't see it), but because younger people, graduates on a student salary, will want to naturally explore, move city with their boyfriend, move on - unfortunately if you train up someone for two years, it's normal that they won't become an early product lead, they will actually leave. Yes maybe you can pay enough for a startup but - experienced people are more likely to stay, and work much better. 1 bad staff member counters 2-3 good ones. At start up level, that's life or death. Try and get HR and management experience (particularly with high churn) before startup so you learn people-HR skills and understand this dynamic, so that you don't kill your startup for someone else extracting from you/your kindness

Secondly, try and do things with 2 magnitudes higher amounts of potential success than you need. Avoid things where you think you can make 10k or even 50k a month - the minimum viable needs to 250k, for part of the market. Otherwise, you'll plateau on either growth, certain marketing channels, employing in the niche - those don't make sense for a sole purpose startup.

Have budget ahead to cover things but set deadlines. If possible, try doing a side hustle/startup with a part time staff member filing corporation taxes 1-2 years ahead. You'll learn the ropes of stuff that will save you valuable time when you need to crush taxes.

Don't let yourself go into CC debt for a company... classic student mistake. generally every $ of credit will take $$ to get out of for a college grad (not just in terms of numbers - you will feel less confident in debt, it'll cost you in other areas of life. Like your partner breaking up with you). Don't listen to peoples advice if it's too general, go for specific advice.

If your results and your self-view is that you are in the top 10% of something and you have received compliments, never take advice on that thing. You'll be told things that apply to the bottom 70% and could make your performance worse... lean on those strengths.

be ready for friends to leave, abandon, disbelieve you, hate that you work late, and don't recognize that they doubted you when you do succeed. Same for partners. Unfortunately, a lot of startup founders start single, and a different group end up single. Doing a startup is a major change for a partnership so consider that before you make the step.

Well inability to own a house was a major prelude to the 1930s economic and later conflict issues, particularly in western Europe (where it was more expensive in places like Britain to buy a house than even today, about 10-14x salary vs 8x today). House prices have been how they are now before: in the 1920s... (America is a bit of an exception because of all the expansion into California etc in the early 1900s).

Male crime rates correlate with economic opportunity too, so even if you don't care about the 1930s and the economic reset of WW2 and its rebuilding (and WW1, but to a lesser extent), you should care about the possibility of crimes rates continuing to regress. Serious crimes don't get solved at a much higher rate than in the past, even if they have decreased since the 90s.

Home ownership is a lagging indicator and doesn't show the whole picture: If people used to buy a home at 25 and move into it, but today they wait and live in their parents house until 30 before moving into a home, the stats might appear the same. The quality of life won't be. Home ownership rate needs context too, in some countries rent is low and house prices are high(er than the US), where many people rent through retirement, and where the implication of rising house prices is not as bad unless rent also raises.

Thank you to author for pointing out some negatives here.

The 5th version made two "non-error", but ergonomic changes, for the sake of larger code bases and standard formats over quick, yet reliable, reactive and testable code (which first attracted me to svelte): - the need for getters and setters when using $runes in functions; this is a waste of time and almost always unnecessary in the contexts in which one uses small store classes. Not that Svelte 4 was better, it was different.. in a pinch, I'd take 4 though. - using folders for routing and using standard page.svelte.js and server.svelte.js naming for everything; it clutters your IDE file list. You could say it's good practise, but limiting to one component per file also causes more files in some cases than multiple components per file (as is possible in React)

Despite all this, Svelte 5 introduces $writable, which is kind of for the use case of one-off simple state - which doesn't need getters/setters - so the whole thing doesn't feel that consistent.

I don't mind using consistent runes - that is a bit of boilerplate, but makes it easier to reason about. I still love the reactivity of using bind (like Vue) and using $effect with it automatically figuring out what to do without the footguns of Reacts useEffect. I love that I can just start writing files and routing will work, css can just go there without a node module to process it.

Increasingly, I've realised React is probably going to remain my number 1: - Recent changes are mostly just shorthands to reduce boilerplate - Mobx has got state down - Hooks have footguns, but I know them - The eco-system is still unmatched - `use` and other statements allowing async loading (like other recent shorthands) is helpful, a lot of my components used to have boilerplate to do async stuff, and it took effort to make failure cases/loading work well. This pairs very well with server functions too.

Yep. And what makes it good is that it automatically zooms in and out when you take actions or move the cursor - it edits for you

I do not know how FastHTML works with AI, but for the recent Svelte 5, which has a similar llms.txt file, it's clear to me that actual usage patterns are required over explanatory llms.txt content with instructions. In Svelte 5 usage, it's consistent at certain things which the docs spell out for version 5 and would not be in the training data (runes) but not in their placement, reactivity, or usage changes implied by the system (e.g. changes of information flow outside of props/states)

It seems similar to a cheat sheet which has formulas, vs a cheat sheet with worked through sample problems that align perfectly with the test.

The latter exists for historic modules - the former is the best you can do for recent libraries/versions.

I am not sure where React stands as I know they have changed their reactivity model and introduced patterns which reduce boilerplate code. Can anyone comment on the latest react version used with AI?

I agree with this though it's only relevant beyond newbie periods. PTs won't talk about it since it's not motivating, which makes sense.

During a stressed part of life, I slid into staying with the same routine, doing 70% intensity (cutting a set or two off each workout, or rushing them), and stopped going up in strength and muscle size. You could say I have bad genetics but I've had some sport talent and been considered "fit" by a few people so I think I'm average/lucky on muscle genetics.

I basically maintained the muscle but I threw away that time. I kept going into the gym 4-5 days a week like I had when putting on gains, but I could have maintained the muscle (as I have since) on 2-3 days a week.

All the difference was in that 70% --> 85% intensity difference. Once I did that, without changing programme, gains came back. I think that this "hidden plateau" is secretly why so many people bail on the gym.

Not in genetics, but the kind of cursory neuroscience education that "AI" courses contain often goes no further than a sample cell at one point in the first weeks. I have found that the majority of AI graduates believe each cell contains one Mitochondria, even those who can explain the chemical processes behind associative memory and understand how there are multiple binding sites/receptors in most of those processes.

It goes to show how much simple diagrams influence understanding. The synaptic gap diagrams show multiple receptors and abstracted neurotransmitters, and gradients of multiple Ca or K ions. The neuron diagram shows one Mitochondria. That start influences their understanding for years.