brainless

I have recently been working with Google Jules and it has a similar approach. It spins up VMs and goes through tasks given.

It does not handle any infrastructure, so no hosting. It allows me to set multiple small tasks, come back and check, confirm and move forward to see a new branch on GitHub. I open a PR, do my checks (locally if I need to) and merge.

I understand that we are slowly taking away our own jobs but I do not find it depressing. I do find it concerning since most people do not talk about this openly. We are not sure how we are restructure so many jobs. If we cannot find jobs, what is the financial future for a large number of people across the world. This needs more thinking, honest acceptance of the situation. It will happen, we should take a positive approach to finding a new future.

Yes the prompts are clear enough but it depends on the capacity of the people involved. People have to internalize the math (or any other) concepts from language into some rules, syntax, etc.

This is what an agent can do with an LLM. LLMs can help take English and generate some sort of an algorithm. The agent stores algorithm not the prompt. I do not know what current commercially available agents do but this was always clear to me.

I am not sure I got your point about English. I thought Karpathy was talking about English being the language of prompts, not output. Outputs can be English but if the goal is to compute using the output, then we need structured output (JSON, snippets of code, etc.), not English.

This is something I have been attempting for quite a while now. One simple tool I started is a deterministic data extraction system where AI helps in finding out the data to be extracted but then the code would try and "template" it. When we have the template, the extraction on any similar string would happen deterministically.

Think of extracting parts of an email subject. LLM is great at going through unseen subject lines and telling us what can be extracted. We ask LLM what it found, where. For things like dates, times, city, country etc, we can then deterministically re-run on new strings to extract.

https://github.com/pixlie/determined

Building a knowledge graph where AI/ML is used to classify nodes (entity type). Next step is to extract data deterministically. https://github.com/pixlie/PixlieAI

There is a crawler to chase given objective and crawl/store web page results. We use Claude Haiku and Brave Search API (own API keys) and we will support local models soon.

Potential use-cases are lead generation, competitor search, product tracking, contracts search.

Hey everyone, we are creating a knowledge graph (runs locally or on cloud), with a simple web crawler, with Anthropic Claude (Ollama & others coming soon). We made good progress this month:

  - Start with an objective, like "track Python/TypeScript jobs from any job site"
  - Anthropic Claude gives starting keywords and crawl guidelines
  - Using Brave Search API, we get URLs
  - Crawl starts

At the moment you can do keyword search. Entity extraction is coming soon, so "job post" will be an extracted entity in the above example and you can browse all jobs from any crawled site in a uniform UI.

https://github.com/pixlie/PixlieAI

Congratulations on the milestone.

I checked their team page, repositories and bounties. It is lovely to see things so well structured around an open source product that encourages everyone to contribute, share, even get paid.

I will keep this in mind for our team calls as a brilliant reference.

I really thought this name was from the common word for "human" in many languages across India. Slightly different pronunciations in some languages but I think many Indians would relate to the word as a derivative of "human".

Hey you, sorry I could not guess you from your username. Thanks a lot.

This sounds lovely. This product has not reached to the point where ontology matters, that is still far away.

I am just going for the low hanging fruits - very basic stuff like location, person, event, activity, date (a few more things) and relations between them. One of the limitations is that I want to do deterministic extraction with suggestions from AI/ML, so there is much code to be written. I will email you, thanks a lot!

I have been working full-time for about 15 months on a product to store real-world entity-relations in a graph (using AI/ML for extraction). The idea is to extract entities and relations deterministically from text (using AI/ML for clues about type and position of entities in text).

It is very much a work in progress with lots of commented out code which are just experiments.

https://github.com/pixlie/PixlieAI

Really lovely homepage and I can personally relate to this. I signed up to check the onboard process.

I am in a similar space and our product creates workflows from prompts. Internally it is a JSON schema generator and response parser (from AI). We focus on creating assistants (personal or team) that reside on local computers, can crawl, read from any data source and build a knowledge graph for search, daily tasks, etc.

In our UI efforts (we are really behind in this) the graph editor is starting to look like how I see in your product.

Thanks, I just searched for Perforce time-lapse. I will find out more about it, more food for thought.

This is such a lovely article and I relate to the idea as I have thought of something similar when I was building a passion project (https://github.com/pixlie/gitplay).

The idea was to use the Git commit log to show how a software evolved into what it is today. I needed to think of zooming in and out of context, where there are thousands of commits involved. I am not a good UI person so I will definitely take inspiration from this.

Python doesn't have a narrow scope. At least compared to Flutter. It's in so many places, from Linux installers, to medical, aerospace, math,... and finally the web (which is probably what we hear about the most).

It will perhaps not be a drop in replacement for vector dbs in every situation but yes it will be so when you want accurate results that follow the semantic chains (entities and relations).

At this moment, we have not entered the territory of conflict resolution but I know what you mean. Interestingly I just came across this: https://arxiv.org/pdf/2410.18415 (released on Oct 24, 2024).

A video explanation (where I originally found the paper): https://www.youtube.com/watch?v=LOF6fvV5XUc

I am building a graph based semantic search engine. We can use low cost LLMs, like Haiku, or local models to extract semantics (named entity recognition).

Then the nodes in the graph maintain types (things like people, date, currency) as extracted and allow queries.

https://github.com/pixlie/PixlieAI

Currently building a demo where we crawl startup investment data to build a knowledge graph that can be filtered for patterns.

The engine can guide the crawl process, to keep crawl limited to the problem statement.

I guess what the other person means is that there is no "buying", you can take the book for free. Yes it is incomplete but you'll get access to future work on the book.

I think there is a whole world between going all in on an open source license and allowing funded companies to simply competing with you vs not releasing source at all.

We need a "fair source", or whatever we call it. We need a way for people to contribute, many ideas to gain a community of contributors, without a funded agent to simply focus on how to monetize the software.

Open source or "fair source" are all distribution models. For people to see, try, adapt and maybe even contribute to your software ideas. It enables everyone to build and share. And if someone wants to compete with the stated software, they can buy a license. That is a fair business model, IMHO.

This looks beautiful. I am building in the same way, a private app, runs on desktop, same tech stack.

I have a very generous free tier and I want to add two paid tiers. I have to figure out a license check that doesn't leak user information.

I would love to collaborate with you and also hear about your future plans for monetization.

Giving you a big hug.

I have been through really rough periods and as my health took a big toll, the only means I had to recovery was to move away from crazy startup life. Gradually I understood that, maybe, there is a better balance. I still want to learn, grow, solve problems, dream big.

Instead of making 3 year plans, I started making 10 year plans. I started taking care of my health, like it is the most important thing. I unplugged for a couple years, lived in cheap places to lower my financial burden.

Now I live in a small village, have built a work routine that has no deadlines and I am happy, very happy. I have a hostel I run here, I write software I love, I am planning a product. I just setup a camping spot, it is lush green around here, a slow and simple life (all shops close at 20:00 for example).

I sacrificed those magazine cover dreams but in return I got a wonderful life and I am building again, just at 0.5x speed. I hesitate to think about accelerators like YC. I know I will panic so much when it comes to all those metrics, money and everything else I cannot process anymore. But that is OK.

It is AGPL that most people will have an issue with. Perhaps you can select a different license for the automatic switch.

I think the most important thing a founder brings is to "not give up". The journey of most startups that innovate is complicated, unknown and unusual. Traditional management is not made for this. I have always felt that "pivot" comes from "founder mode" itself. It is one way of keep trying, even when a founder accepts that the original idea is not working well. Many founders have attempted and failed multiple times to finally get to something that works. This value is less visible in any other domain of management.

As a solo founder trying again (1) after many failed attempts, I can only say that my investment in Rust is paying off. A lot.

I have worked with PHP in my early days (2006...), for about 6 years. I wanted to move to a language more structured. When I switched to Python I felt it was worth my investment. Working as a web developer, JavaScript was always there and I could not change that. After working with Python for 11 years, I felt if I attempt to be a founder again, I needed two things from a language - structure in developing my ideas, peace of mind to take a product to users' hands. I was OK to learn syntax, but I can not become a legendary developer with magical knowledge of some language.

Refactoring the codebase is wildly different since I get so much support from the compiler/type system. And, if it compiles, it will simply not have a class of errors.

The investment is steep, really steep. I am no distinguished engineer, just another one with average talent. I have been working full-time building my product and I am preparing to launch - multiple products in the next 6-7 months, all on the same core code-base in Rust. I have never felt a language could give me this much.

1. https://github.com/brainless/dwata

I have been working full-time since early 2024 on a locally running app, where I stick to open file formats and SQLite for storage. I would like to make this "local-first" - I feel I'm already on the way.

It is an open source app to download pervade data and provide fast search, offline access, automatic labels, categories, summaries, etc

Starting with emails, calendars and attachments. Integrations for Slack, LinkedIn, Stripe, databases, even web crawl are coming...

I plan to have real-time collaboration on top to make money (ideally with end to end encryption).

I take privacy very seriously. I do not even have OAuth2 apps for Google to access your data - you have to create those yourself (there is documentation in the app).

This branch has updated readme:

https://github.com/brainless/dwata/tree/feature/prepare_mvp_...

Open source app to download emails, provide fast search, offline access, automatic labels, categories, summaries, surface up what needs our attention, etc. (some features need AI).

Starting with emails, calendars and attachments. Integrations for Slack, LinkedIn, Stripe, databases, even web crawl are coming...

This branch has updated readme:

https://github.com/brainless/dwata/tree/feature/prepare_mvp_...

As a child, I used to spend a lot of time with insects. I wanted to understand how they understand the world. I have thought about death and how other animals may perceive it. In the recent years, with a lot of machine comprehension, I find myself asking these questions again. And I spend a lot more time with animals now, so that aids in the process.

In a way a deeper question I have is do we humans actually "understand" death in similar ways? I personally do not think so and it arises from how much or little I have seen people spend time pondering on these topics. With some, I can discuss hypotheticals easily and with others I just can not. They do not admire these subjects and I wonder if our brains just perceive some of these things differently. Time is another such topic.

I should add that we do not need to agree on what is "death" in order to grieve the missing. But I guess how would cats know death vs missing (vacation)?

Congrats on the launch and kudos on building on bold ideas.

I think car rides are "good enough" for a certain distance vs short flights. Short flights (tiny aircrafts) are so expensive because they have to compete with the good enough option and I am not sure that can ever happen.

The more important point is that flight accidents can affect people or buildings that are far away from roads - practically anywhere within a certain range of a flight path. This will have huge effect in people's mindset about safety and work against the popularity of such flights.