iraldir

Amazing model at this size if true, that's quite crazy!

Really cool and love the design, will give it a go

I'd personally target the "source" of the addictiveness: outlaw behavioral analytics that, on those apps, analyse any detail of your navigation (how fast you scroll, where you stop etc.).

A: Many people are not aware that's a thing: "I did not click like on a racist video, but it keeps showing them to me", when in reality, the algorithm detected that you slowed down when shown that type of content so decided to feed more of it to you.

B: I targets your worse instinct (oh a half naked girl, maybe she has something interesting to say?), and if you're not aware of point A, it's a loosing battle. If you are aware of point A, it's constantly tiring to have to adjust you speed scroll etc.

I would personally make a GDPR type law that would require: "Any feed of user data must be presented in a non-opaque way, such as chronological, ranked by upvotes etc. Any personalized feed based on expectations of user preference should be entirely transparent, by allowing the user to access, modify their profile, as well as locking them (preventing future behaviour from changing the type of content you see).

Interesting article (did not finish it), however the little apartes feel soooo condescending. It's impeding the reading in my case. Might just be me though, don't pay too much of a mind to it.

The overruling prompt of an LLM is essentially "give the most likely answer to the text above".

If you ask an LLM to say left 80% of the time and right 20%, then "the most likely answer to the text above" is left 100% of the time.

It's actually a specialty from belgium / north of France called speculoos, biscoff being just a brand of that type of biscuits that managed to export them worldwide.

In the north of France where I lived for a while, it's an absolute staple akin to what peanut butter is to an american maybe or Matcha to a Japanese. Speculoos butter is spread on bread, lots of pastries are speculoos flavoured etc.

The market is really tough right now. I'm just starting to look for full-stack positions in London, this time last year I was receiving lots of offers constantly, but now it's a very different story.

I've so far only contacted one recruiter I trust, and even he only had one job I could apply to.

BTW, if anyone is looking for a full stack web dev in London (TypeScript, React, Next, Node / Bun / Deno, even a bit of Rust), you can contact me at https://www.lajili.com).

Oh boy did that send me down a trip to memory lane. Played it as a kid in France, probably around 2002? God knows where I found that.

My 2 years 5 month old toddler uses Adobe fresco on my wacom tablet every now and then, she knows how to switch colours and draw, I erase the page every now and then for her.

She also watches every now and then "C'est pas sorcier", a french documentary series for kids that's pretty advanced (go into details of biology etc.), on all sorts of topics (nature, science, food etc.). That's also a good way for her to experience some French as we live in the UK and her only source of French are talking to me, or to her grandmother on the phone.

Just to be clear she does plenty of other activities, we cook together (she has a children's chef knife and uses it to cut things like tofu, carrots etc.), she draws on paper, she sings and dance, goes to the swimming pool weekly etc.

As for video game, that's more for later as she does not have yet the skills for that but at a glance

- Logic : Puzzle games, things like incredible machines, but also management game like sim-city. - Creativity : Any sandbox game, the sims, minecraft, drawing game, sculpting game, animal crossing etc. - Motor skills: Platformers, fighting games...

Essentially any game that requires a lot of effort for little reward, as opposed to games that makes you touch something shiny and shower you in visual and auditive feedback.

Yeah for sure, MLP friendship is magic is there in the list of high quality show. It's unavoidable that toys are sold on popular francises, I just don't want show that are obviously just 20 minutes long advertisements with little substance.

As always with that sort of conversation, I don't think "screen" is a valid category of interaction. Drawing on photoshop, watching a documentary on whales, playing bejeweled or using TikTok are very different activities that impact the brain differently.

My daughter is still young (2.5 years old), but I know that I'll let her play video games when she wants as long as it's as part of other activities in the day and that the games are ones that develop either her logic, creativity or motor skills.

Similarly, watching a quality show like bluey, or watching a full movie end to finish like Totoro in Japanese are watching activities I support, whereas watching some crap cartoon made to sell toys is not.

As always with GPT it's a hard balancing act. There is probably an instruction to ensure it does not visit urls un-necessarily, otherwise every request out there would trigger a search on bing if only to get more up to date information. Like "what is the capital of France" "Well let me ask Bing, you never know". Whatever protection they have is clearly overcorrecting in this instance.

Another point to support the same idea is data can be falacious.

For instance, you are making a pretty advanced 3D web app, and notice in your analytics that your userbase is only chrome and safari users.

An easy conclusion is to focus your testing on those two platforms, or maybe even drop support entirely for Firefox and Edge by using some webkit specific API.

A not so easy conclusion is the experience might be so bad or buggy on a non-webkit browser that anyone who tries the app in those just gives up on it.

The reasonable truth in this case? You should use standard browser distribution except if you're operating in a specific market, it might also be perfectly fine to drop non webkit browser if the ROI of developing them is not worth it for your goal etc. All of which does not need data but rather intuition and common sense.

- Did not work on firefox for me (start conversation and nothing would happen, I would not hear the voice) - on chrome it would not allow to change my microphone so had to open my macbook. - Also on firefox, when I first logged in, clicking on he "try now" button would send me back to the landing page, I had to go and click on the playground

With that out of the way, it's really interesting. The challenge I suppose is how narrow / wide the API made by the developer is. A narrow case, like "booking a dentist appointment", might feel like a step down compared to an online form, and most likely would fail short of satisfying someone calling, because if they're calling they had some deeper needs.

On the other hand, a wide API, like if you gave access to pricing information of the dental practice, info about the doctors, a way to reach an actual human being, health advice and post op advice, generate documents from past appointments etc, and you have a higher chance of hallucination, misclassification of what the user wants etc.

I'm still not sure if the best approach is just to leave the AI to deal with that, or to represent the user intent etc. with some sort of State Machine. Maybe a case by case etc. But the more you add in that logic, the slower the machine is.

New version of caméra café (https://en.wikipedia.org/wiki/Cam%C3%A9ra_Caf%C3%A9) soon on Amazon Prime?

I was going to be fairly underwhelmed, I mean, yeah it's a web game of course you can play it on its website, wooplidoo.

But actually having a demo directly in the middle of the action embeded in an iframe of the landing page is quite neat. It blends properly with the rest as well, so actually well done.

Some feedback: - Demo plays music. If the user clicks the purchase button, a new page opens which also plays music, this can be quite unpleasant as they are both playing at the same time. - Personally I like the art of the game, including the pixel art character faces, but I don't like the banner. The banner feels like it was drawn by a teenager discovering photoshop with a book "learn to draw manga style in 2 hours" – That might be a bit unfair and I'm sure it was a lot of effort, but a lot of details make it feel amateur-ish in the wrong way, in particular the most central character and its face. Have you considered using the same pixelised style that you use for describing each character bellow? - on firefox the character faces overlap with the text on "Create your own maps..." (Firefox, Mac OS, very large screen)

I find the comments on HN a bit harsh (at the time of writing anyway).

The workday as described in this article (8am to 2:30pm) does leave plenty of time to spend with family, friends and on hobbies, even on weekdays.

I don't have full control of my schedule at the minute, but if I did this could be quite interesting

For personal projects, I really wish there was some sort of semantic CSS with many different implementations, which this could be one of them.

What I mean is a set of rules on how to structure your HTML (like use a main element here) with some set expectations of how it will be structured on the page.

Then, many different people could write different spritesheet that makes the same HTML looks widely different. I'm not just talking colour, but fonts, radius, opacity, animations, etc. etc.

One stylesheet could make the HTML look barebone black and white, where another makes it look all in gradients of purple with fancy animations etc.

All for the purpose of writing simple application where you don't care so much about how they look.

- classes would not be used - CSS variables would not be used - just plain old div, H1, section etc.

Hi everyone, just wanted to share this project of mine, I shared it a long time ago in here whilst it was just a landing page, now the full product is available in Beta and everything is free for the duration of the beta.

ChartMyLife.ai simplifies diary keeping and life tracking. Leveraging OpenAI's GPT-4, our app turns your spoken words into organized diary entries, tracks key metrics like diet and exercise, and offers personalized quests and insights from AI coaches. It's a straightforward tool for self-improvement and staying on top of your goals.

Now for the nerds of HN, some more technical stuff

- ChartMyLife uses GPT-4 and whisper to let you talk about your day and store it in a structured way. This means both a text based diary (today I did so and so) and tracker entries (Lunch: 560 calories etc.)

- A second set of AI prompts review the diary weekly and can give you "quests" with an experience point system. The AI judges how much each quest is worth based on a barem it was given

- Dalle-3 is used to generate Quest icons

- This is made using NextJS with app router, server actions and cron jobs. Entirely serverless, and running with mongodb for the DB. The biggest challenge with this architecture is dealing with dates, as server side rendered component might run on UTC whereas component ran on the user's device are in a local time zone.

- I'm using JoyUI (from the makers of MaterialUI) which has been mostly a nice experience.

- One cool tidbit is I have AI tests to test my prompts. They look like this:

- All the illustrations are generated by dalle-3 or Midjourney V5.

``` it('knows how to extract a meal and a workout', async () => { const result = await aiPrompt("I ate 6 nigiri sushis and I ran 5km", simpleUser, "23/11/2023, 12:08:58");

        const validation = await aiValidate(
          result,
          'It should create one entry in the meal tracker for 6 sushis, and an entry in the gym sessions tracker for 5km. It should not create any other entries.'
        )
      });

``` An AI is used to validate the result so it's the ultimate integration test that is easy to write. Each test is run 10 times as some prompts have a tendency to work some times and not some other times. I'm looking at 90% success for now to consider a prompt successfull, though most of them are 100%. It's ran on CI only when changes happen in the file related to prompting openAI.

Mine says differently: https://chat.openai.com/share/792e3964-f96d-47c6-a598-b44791...

You cannot trust what it says about itself. As for knowing catastrophe, that's a pretty poor example as open ai as always it keeps the model up to date with major events (war in ukraine for instance).

Padding stripped, thanks for your feedback

That's because you're building on an existing consciouness. Even if you wipe their short term memory, you still have their long term memory from before the whole sensory deprivation tank that have formed concepts such as the self, the world etc.

I see that you're seeing the training phase of GPT as the equivalent time the adult had before, but I don't think that works. Part of what makes a consciousness emerges for me is understanding your place in the world, how your actions are having consequences etc. Being nice can have people be nice to you in return, or depending on the environment, you can learn that by being insuffurable people will just give you whatever you want (cf some toddlers). GPT's training comes short of that. If you had an actual session of GPT connected to some sort of long term vector database where it can store what it learns based on conversations, to differenciate individuals and itself etc. that might give it a fighting chance to develop an equivalent of a conscience.

That's explored pretty well in science fiction, but I guess for me an easy reasoning is if it's feasible, then someone will inevitably do it. If so, then you might as well be the one in question, this way - you get fame and glory, even if the results are disatrous (wasn't there a blockbuster recently about the father of the atomic bomb?) - if you intend well and you believe in your own capacity, you might prefer creating your own AI than waiting for someone with a different agenda to do them - just... curiosity? Loneliness for some? - if you believe the AI will discriminate against people who didn't help it, you might want to be on its good side?

I'm spitballing here but surely given the number of people on the planet, you'll find someone with both the skills and the want to try it

Sorry I did my review / typo correcting on the text editor of languagetool... which for some reason does not support "Command+C" in the editor, only right click copy. So my copy failed and I pasted the same error filled text I started with

Should be corrected in a few seconds whilst CI does its job

Current POC is ugly and made for my own personal use. I'd be inclined to start spending time making it look decent for a screen record etc. but I'm trying to go against that instinct and to focus on user acquisition instead.

If I can't get any emails from the landing page, then I might as well just keep using the POC for myself and not spend days, quickly turning into months worth of time and energy, into making it into a "product" when maybe there simply isn't a market for it.

The wikipedia page listed in another comment makes the point that this also (can) use three meanings, as Police is also a place in Poland.

The police police from Police Poland police the police from Police Poland: Police police police police Police police