marc

Over the years I noticed many of my friends collect all kinds of different photos.

One takes a photo whenever he sees an exotic car to share it in a group chat. Another told me and his girlfriend have a passion for traffic cones (apparently there are many different types).

I thought it would be fun to build them an app to collect all of these. And of course, with modern technology, we can cut out the object (i.e. remove the background), automatically identify metadata such as the model, colors, type, etc, and put them on a map to find back later.

That's what Loot is. It's my second ever iPhone app (I built many websites before) and it's completely vibe coded. But I did put in A LOT of effort in not making it look vibe coded. A lot of time was spent polishing the UX to give it character and make it fun to use.

The interactivity of the website gives a good preview of the playfulness of the app.

I already got a ton of good feedback and would love to hear what HN thinks of it and how you could see yourself using it. What features you'd like to see added next, etc.

It's a free download with currently no in-app purchases or subscriptions.

For many years we explicitly denied any startup that was already launched.

This was back in the day where most startups had a long "pre-launch" phase where they would be collecting user feedback through private betas, etc before launching publicly.

I wanted BetaList to be focused on early-stage startups and to avoid a lot of debates about what is and isn't "early-stage" I decided on that clear policy.

We also denied startups for other reasons such as using an over-used homepage template (they never converted well for our audience as they'd see the same one multiple times a week) or simply having a dime-in-a-dozen product (which also didn't get many signups when featured on BetaList).

Haha you're welcome. Thanks for letting us share it with our audience :)

Generally you can just reach out via a contact form. I should link to those

Yeah PH launched a few years after us :)

Fun story:

I built BetaList 16 years ago which was one of the first "product discovery" platforms. Years before Product Hunt, etc.

I manually reviewed every submission and unfortunately often I had to tell founders that their startup didn't qualify to be included. Almost everyone would (understandably!) argue their case, but as volume increased I couldn't afford to go into a deep argument with every single founder.

That's when I made https://submit.co a site similar to OP's. The idea being that instead of say "No, we will not feature your startup" I now gave them an alternative place to put their energy.

Initially it was mostly a list of tech blogs, but as more product discovery platforms popped up, I started adding them too. In a sense, I was promoting my competition but it was exactly the startups we couldn't list any way for one reason or another.

Eventually that list of "places to submit your startup" got so popular (and copied everywhere ) that it started driving traffic back to BetaList. (I included it at the very top of the list).

The reason new social networks rarely adopt domain-based handles is that usernames are one of the few forms of “currency” they have.

Scarce handles create urgency. People rush to sign up and claim their first name, etc. With domain-based handles, that lever disappears.

It’s one of the most common growth tactics I see by social networks launching on BetaList.

I’ve had this idea for many years, but never found the time to build it. Now, with AI, it only took about an hour to build.

The objective of the game is to put all arrows in alignment. Point them the same way.

When you click an arrow, it rotates 90° clockwise. It also rotates all neighboring arrows.

All puzzles are solvable, but as you increase the grid size and difficulty level it will become progressively harder.

Please give it a try and let me know what you think. I’d also love to hear any strategies you come up with.

OpenAI doesn’t take security seriously.

I reported a vulnerability to them that allowed you to get IP addresses of their paying customers.

OpenAI responded “Not applicable” indicating they don’t think it was a serious issue.

The PoC was very easy to understand and simple to replicate.

Edit: I guess I might as well disclose it here since they don’t consider it an issue. They were/are(?) hot linking logo images of third-party plugins. When you open their plugin store it loads a couple dozen of them instantly. This allows those plugin developers (of which there are many) to track the IP addresses and possibly more of who made these requests. It’s straight forward to become a plugin developer and get included. IP tracking is invisible to the user and OpenAI. A simple fix is to proxy these images and/or cache them on the OpenAI server.

Can you share more about your OpenObserve setup?

I’d love to know:

1) How it compares to Sentry in terms of insights you get 2) How you set it up as an accessory

I’ve used various (hosted) APM services for my Rails apps but they all are stupendously expensive so your approach sounds intriguing.

Hope it’s okay I shared it here. Saw it in my RSS reader and figured HN would enjoy it.

Glad to hear it served as inspiration. The video looks promising!

I had my Apple Pro Display XDR stolen from the office in Lisbon, Portugal.

They also stole my AirPods Max which has Find My technology built in.

It happened over the weekend and when I found out about the burglary about 2 days later I checked the Find My app. I saw my AirPods Max was in a small town about 3 hours away from my office.

I informed the police and they said there’s nothing they could do about it due to privacy concerns. Ugh.

I considered going there myself, but it being in a foreign country, not knowing the language, having no transportation, and not knowing what or who I’d find when I arrive, I ultimately decided to not risk it and never ended up going.

A few days later the signal went offline. Months after that it showed up on the map again. It was in Moldova at a shopping mall known for selling electronic parts.

I don’t know what happened to my Apple Pro Display XDR, but if you ever see one for sale in Portugal please send me a message :D

BTW, interesting side story: They didn’t just steal my display and headphones, but also many computers from other offices. Multiple floors except one. I guess they didn’t have time for that one. A week later they came back to rob that other floor as well.

I discovered emoji .to domains ~7 years ago and put up a site listing all the available ones [1]

Within a few days almost all of them sold and made a couple grand in affiliate commissions.

I wrote about it here: https://marc.io/emoji-domains

Email forwarding is also a clever use! Nice to get that recurring revenue.

[1] https://xn--f28h.to/

I don't have a Vision Pro yet (I'm outside US), was curious to see what apps are being developed. So I made a list!

I spend a lot of time on making it easy to navigate and if you add it your home screen it almost feels like a native app.

Tech stack: Ruby on Rails + Tailwind CSS + Hotwire

Last year I made the first Stable Diffusion app [1] to redesign your interior (upload a picture of your room, and it restyles it to a different design style).

Last week, GPT-4 Vision was released [2]. It allows developers to upload a picture and ask questions about it.

I figured it would be be a compelling product to get a interior design consultation, by AI.

The feedback is surprisingly detailed and useful. It understand the layout of the room and can give actionable recommendations on how to improve the flow, overall aesthetics, etc.

For people on a budget, it might be a great starting point.

The next step will be to visualize the recommendations with AI (Stable Diffusion) so you get a better sense of what your room could look like.

Curious to get your thoughts!

[1] https://roomai.com [2] https://platform.openai.com/docs/guides/vision

Arnold kept inspiring people from all walks of life that through hard work you can, in the case of body building quite literally, shape your own reality.

I’d call that being useful.

I think this is based on https://replicate.com/fofr/sdxl-emoji

It's SDXL (Stable Diffusion's latest release), fine-tuned on Apple Emoji images.

Great write-up!

I appreciate that you went the extra mile by showing how a generic enough controller can be used for various use cases.

People coming a from traditional JS background tend to make very specific Stimulus controllers, but when you make them a bit more abstract and modular they become more usable and easier to maintain.

Assume the end-user can write arbitrary SQL and LLM is just an interface for that.

Creating a read-only Postgres user with limited access might be a good workaround.

Not sure about avoiding infinite loops, CPU loads, etc. Curious to get an expert’s input on this.

Not sure why I'm being downvoted, but here's my reasoning:

"Apple being granted the additional rights they seek (unlikely)"

Trademark tend to be narrowly scoped to the products and services being marketed by the company.

"them then wanting to enforce their mark against this fruit company (unlikely)"

They have more to lose (reputation) than to gain.

"winning litigation (unlikely)"

There's no likelihood of confusion. Dictionary words can be trademarked, but generally not for the goods and services they describe. Fruit company has used the mark before Apple even existed. Trademark rights are based on first use, regardless of filing.

The key word here is “could”.

It assumes a lot of conditions will be met including Apple being granted the additional rights they seek (unlikely), them then wanting to enforce their mark against this fruit company (unlikely), and winning litigation (unlikely).

TL;DR click bait

Looks great. What type of feedback are you looking for? [1]

Also curious what got you interested in building this specific product?

Either way seems like a great start

[1] https://marc.io/ask-for-feedback

I still struggle with this, but I’ve become better at it.

What I learned is that “the last 10%” really is about half of the work. The solution is to reduce the scope. A lot.

Not only will it make it easier to finish the project. It also allows you to get real-world feedback a lot earlier. Which is both rewarding and informing. Two things that will help you push your project forward if you choose to do so.

In case you have your eye on an inactive Twitter handle, now is a good time to monitor it.

Shameless plug: https://handle.horse let’s your monitor Twitter handles for availability (has a free plan)

Ultimately I don't think it matters whether it was written by ChatGPT or not.

If it conveys the message the OP is trying to communicate in a readable way, that's all that matters.

I think the reason the person above me pointed out it seemed like ChatGPT (as did I), is to inform the OP that it's quite transparent and they shouldn't just take whatever ChatGPT spits out.

Of course it's possible that we're wrong and it was manually written. In that case it's useful to realize we live in a post-ChatGPT world and if something _looks_ AI-generated, that by itself is (unfortunately?) a bad attribute now.

OK, before you post “who would ever listen to AI-generated podcasts?” please listen to a few. I think you’ll be impressed with the voice quality.

What excites me about AI-generated podcasts is not to produce a lot of mediocre quality at scale. But that for the first time ever it becomes economically viable to generate audio content personalized to your individual preferences.

I’d love to get your suggestions. But right now I’m considering the following types of data sources:

- Local weather forecast - Daily email summary - News headlines for your specific industry - Updates on your (stock) portfolio companies

The product isn’t quite there yet. But one idea is to provide you with a daily morning episode just for you.

Please let me know what you think.

I love that you wrote it using Cloudflare workers and just kept the default URL. Not everything needs to be a branded product with a custom domain, logo, etc.

When you use ChatGPT a lot you start to recognize certain patterns.

In this case, there’s 3 paragraphs of text, but there isn’t 3 paragraphs worth of information.

So what likely happened is that the OP provided limited context to the AI and asked it to write a 3-paragraph post based on it which it happily obliged to.

It feels like GPT-3.5. I think GPT-4 generally does a better job at these types of texts.

For people who are looking for a hosted solution: https://t.me/marcbot

Being able to use voice messages as an interface makes a huge difference. I can just ramble on, sharing my thoughts, and then have GPT turn it into something sensible.

Great for brainstorming, getting your thoughts out on "paper", etc.

I made https://t.me/marcbot a few weeks ago. Comes with voice memo support, etc.