yamalight

I've started self-hosting email in ~2021. It took about 6 months (of mostly waiting) to clean up reputation of IP I've gotten (was in a few spam lists). After that - it was incredibly easy to maintain. It's literally just running mailcow update every now and then and that's it. So, if you are willing to fight for a bit in that initial stretch - no, it's not hard; yes, you can self-host your email.

Not usually, no. (edit: and that totally went over my head, lol. good one :) )

It told me my ~10 year old js project was 50% AI generated. Yeah, this is more or less the same as "AI text detector" stuff that won't work reliably (but people who don't understand LLMs will still use it to blame others)

Those games are usually brilliant - but those are very rare. Like "once in a few years" kind of rare IMO. While that is a valid approach, I play way more than that haha!

What I found interesting with Vaporlens is that it surfaces things that people think about the game - and if you find games where you like all the positives and don't mind largest negatives (because those are very often very subjective) - you're in a for a pretty good time.

It's also quite amusing to me that using fairly basic vector similarity on points text resulted in a pretty decent "similar games" section :D

Built vaporlens.app in my free time using LLMs (specifically gemini, first 2.0-flash, recently moved to 2.5-flash).

It processes Steam game reviews and provides one page summary of what people thing about the game. Have been gradually improving it and adding some features from community feedback. Has been good fun.

I've started self-hosting mailcow (dockerized) on a Hetzner VM ~6 years ago or so. It took about 6 month to clean the rep of my IP address as it unfortunately been in spam lists previously - but after that, it's been a pretty smooth sailing. And all for a measly ~10 euro/mo including daily backups.

Having a research background - for me, doing this in research was always easier. Because you generally have very straightforward ways to validate your hypotheses.

Validating things with customers - in my experience - can be extremely tricky as they might not even know what they want

To finally figure out how to effectively design and test MVPs / prototypes without overdoing it. I think this has been a bane of my startup life for pretty much whole ~14 years I've been doing it.

I really like Edge's "site access: on click" for extensions. Hoping for FF to add it at some point

Happy to clear the misunderstanding :)

We're using top level wikidata classes (you can see specific classes in JSON response). Full list is not published yet, but will be available in the near future.

So, er, how exactly would I prove it? Do you want me to share our code / db / etc we've worked on for past ~two years? :)

For one - how many wikidata classes exactly do you get from Wikineural? If I remember correctly, it can do four (person, location, organization, other). Our models do several thousands.

It'll likely annotate similar things in text since our model is also transformers-based (which is basically current state of art) - can't really do anything about that.

edit: phrasing.

Same with Wikineural - it's a great project, but falls under "prior art". We have our own custom datasets / models / code for NER as well.

That is something we'll be adding shortly. It's not there yet since we weren't quite ready to go public yet.

Thanks for feedback! I'll indeed be adjusting examples to always return results as was suggested here.

We have our own custom datasets, models and code we've used to train them. REBEL can be considered "prior art" though :)

Yep, that'd be the pitch! In the beginning we'll just provide API for people who know what they need / want. Later on the plan is to have "all-in-one" products for end users directly (but that'll take time).

On KGs and industry - as far as we are aware, they are quite widespread. Most of fortune 500 companies use KGs in some form. QA is definitely one of the applications. There's also been quite a bit of work done on e.g. explainable AI using KGs lately (one of the areas we're working on as well).

I've lowered the rate-limiting to 1s, so it should be less of an issue now. But I like the idea of adding examples to exceptions, thanks!

Thank you :D

Thank you! :)

Named entity recognition is typically used to locate and classify named entities in text. So you'd want to have a text that mentions specific things - companies, people, locations, etc. Abstract things like "account" or "subscription" don't technically fall under "named entities" category.

Hi, CTO of DataBorg here. Thanks for trying it out! And apologies for the mess - we weren't quite ready to go public just yet :)

Current rate-limiting is IP based, so it might be your shared IP public address messing things up. The next update we're rolling out over the next few days should make it less aggressive.

If you login with your github / email - you should be able to try thing out without rate-limiting issues. And if 300 credits is too little - feel free to reach out to me at tim at databorg.ai - I'll set you up with a month of free Hobby tier (that'll be adding soon) :)

Hi, CTO of DataBorg here. Thanks for trying it out! And apologies for the mess - we weren't quite ready to go public just yet :)

Current rate-limiting is IP based, so it might be your shared IP public address messing things up. The next update we're rolling out over the next few days should make it less aggressive.

If you login with your github / email - you should be able to try thing out without rate-limiting issues. And if 300 credits is too little - feel free to reach out to me at tim at databorg.ai - I'll set you up with a month of free Hobby tier (that'll be adding soon) :)

Hi, CTO of DataBorg here. Thanks for trying it out! We weren't quite ready to announce it to public just yet :) But hey, can't do much about it now. Pricing is still a bit of placeholder - there'll be 10x more credits on free tier soon-ish.

There is quite a number of ways you could utilize named entity recognition (NER) and/or knowledge graphs (KGs). Ranging from extracting mentioned entities (to e.g. provide a quick access to all articles containing specific entity), to semantic search, to building a unified knowledge graph from text (unstructured) data you have. Cool thing about KGs is that they are based on open standards, so once you've built them out of the data you have - there's quite a few existing tools that (for the most part) work out-of-the-box with them.

Hi, CTO of DataBorg here. Thanks for trying it out! We weren't quite ready to announce it to public just yet :) But hey, can't do much about it now.

Pricing is still a placeholder basically. We want to be in line with industry (which is generally ~0.001$ per 1000 characters), so the final tiers would look something like this:

Free - 3,000 credits

Hobby - 50,000 credits / 49$

Pro - 300,000 credits / 299$

Business - 5,000,000 credits / 4999$

If you could email me privately at tim at databorg.ai, I could give you a free month of hobby tier as apologies for this mess :)

edit: formatting

1. Since there are plans to open source - would be very interested to see benchmarks published alongside code!

2-3. Got it, thanks!

That looks pretty neat, although not all things are exactly clear.

1. You claim that existing graph databases were not fast enough - do you have any benchmark data that compares them with your solution on given dataset?

2. From the description - it seems like you are focusing purely on Person type of data - is that correct? Or is that just the first use case / demo?

3. Do you support more advanced query langs, e.g. SPARQL?

edit: formatting

New version of Exoframe - https://github.com/exoframejs/exoframe

It's a self-hosted one command deployment tool that makes running CD to your own VPS quite trivial.

Current version allows to deploy any dockerized apps quite easily, but I really wanted to have a simple way to deploy Node.js functions (be it HTTP, background process, or trigger/reaction). So Exoframe v5 includes is exactly that (and nearly ready!).

Create index.js, run `exoframe init -f` and then `exoframe` is all it'll take to deploy a function once I'm done. I'm quite happy with the result :)

I always look for articles that talk about issues that arised with that specific tech during / after development. Edge cases and workarounds, things you cannot do at all, maintenance, etc - this is IMO the most interesting bits when you are evaluating a new technology.

Have been rolling my own server on Scaleway / Hetzner and deploying with Exoframe [1] for past ~year. Works pretty well :)

[1] https://github.com/exoframejs/exoframe