I am planning out (nothing to show publicly yet) an automated dynamic algorithm configuration framework for solvers of difficult optimization problems. It is cool because there is a bunch of techniques from all over different kinda isolated areas of academia which became more accessible to hobbyist due to the covid era. There is also a lot of beauty in some results i plan to use. https://www.researchgate.net/publication/3516098_Optimal_spe... .
I am also working on letting computers search algorithms that lets one multiply square matrices asymptotically quicker. Whether i will find any method that will yield any improvement over the state of the art i do not know yet. But it is cool because it is kinda like a puzzle where i can let computers chew on it for a while before come back to it later and have a look whether i see a way to improve the process. If someone has a lot of surplus compute renewable powered i could use occasionally i wouldn't mind such an offer. The problem is really embarrassingly parallel and probably also resumeable in case of power off.
It is at present just at conceptual level, but I am talking to the stakeholders.
Why is this cool? Well, if this works, then not only will it make education easily accessible to everyone but it will also fundamentally change how education benefits society.
Going to plug it and consider this my weekly marketing complete so
that I can justify going back to coding.
I'm building https://feetr.io which is a stock discovery algorithm and
it's cool because it found two stocks yesterday: AMC and BBBY. AMC
went up 34% and BBBY went up 24%.
What makes it cooler, I think, is that it is averaging 4.2% daily
increase across all stocks, with the best performing of each day
averaging 6%.
What might make it cool for the HN crowd is that every single part of
it is written in common lisp. I got a lot of "are you sure" at the
beginning because there is this entrenched idea that lisp can't be
used for serious work. Happy to report that that view is completely
wrong in my case.
I get so excited to build each individual component and so I don't
suffer from the burn out or flailing interest that seems to plague a
lot of other indie hackers.
Feetr will be a paid service when it launches but I think very
justifiable at $8.99 per month. As it hasn't launched, I'm spending
time talking through the stocks that it's watching during premarket
trading, hoping to show people how it works, what it's looking for and
potentially what to expect for the coming day.
Matano - (https://github.com/matanolabs/matano), an open source security lake platform on AWS. Let's you run security log analytics on an open data lake, and detections as code in your AWS account.
Cool because it's all OSS, serverless, high scale, low cost, and built with modern technologies like Rust and Apache Iceberg.
An alternative Microsoft Teams client. It's cool because it doesn't warm your computer as the official client does. (https://www.opercom.co.uk/) the website needs a complete revamping and the UI is being changed to be more modern.
It brings the stock market to the blockchain. Like money (USDC, USDT) you bring into the blockchain it needs to be represented by some token, so when you buy stock you get token(AAPL token when buying Apple) with the shares you bought. You own the shares, get dividends and voting rights.
First version is only buy and sell, next is to allow services to use the smart contract, allowing defi services to provide not only crypto tokens on their exchanges but also stocks.
There are currently over 5.600 different companies available for purchase
Currently it is in JavaScript since it is the easiest to emulate in the browser (I just run the code in a Worker instance, then display the results back to the user).
Kinda ideal for moving from one source to one table, I'm still figuring out how it should work if the user wants to combine data from multiple sources.
I find the JavaScript code not very useful for human input, if the intended audience is not restricted to web developers (a code like interface might still be a good idea). If your intended offering is for devs it would neat some more advanced features to be useful.
More interesting is the question whether you want to support joins. If you do not then you are restricted to concatenation (either in length or in width). Another question is do you want to allow filtering too?
If you want joins i would recommend writing a small query planner because there are some hugely performant optimisations with joins (and filters). If you restrict the possible calculations across loops to simple things like sums, max one could parallelize it between workers with some workm
An interesting avenue for small scale monetisation would be to allow paying users to run them "in the cloud" in JavaScript edge VMs (not sure how they are called exactly cloudflare and others have them). One premium feature could be to pick one of multiple locations where to start them between closest to the data sources or closest data consumer based on the expected ratio of input/output. For large local files and poor connections client side calculations could be used automatically.
https://github.com/hofstadter-io/hof - data modeling & code gen tool, open-source and open-ecosystem. The goal is to code gen the majority (80%) of the mundane code so we only have to work on the interesting parts. You can use any language or technology.
I am planning out (nothing to show publicly yet) an automated dynamic algorithm configuration framework for solvers of difficult optimization problems. It is cool because there is a bunch of techniques from all over different kinda isolated areas of academia which became more accessible to hobbyist due to the covid era. There is also a lot of beauty in some results i plan to use. https://www.researchgate.net/publication/3516098_Optimal_spe... .
I am also working on letting computers search algorithms that lets one multiply square matrices asymptotically quicker. Whether i will find any method that will yield any improvement over the state of the art i do not know yet. But it is cool because it is kinda like a puzzle where i can let computers chew on it for a while before come back to it later and have a look whether i see a way to improve the process. If someone has a lot of surplus compute renewable powered i could use occasionally i wouldn't mind such an offer. The problem is really embarrassingly parallel and probably also resumeable in case of power off.
I am working on https://loan-free-ed.neocities.org
It is at present just at conceptual level, but I am talking to the stakeholders.
Why is this cool? Well, if this works, then not only will it make education easily accessible to everyone but it will also fundamentally change how education benefits society.
Feedback and collaborators welcome :)
Going to plug it and consider this my weekly marketing complete so that I can justify going back to coding.
I'm building https://feetr.io which is a stock discovery algorithm and it's cool because it found two stocks yesterday: AMC and BBBY. AMC went up 34% and BBBY went up 24%.
What makes it cooler, I think, is that it is averaging 4.2% daily increase across all stocks, with the best performing of each day averaging 6%.
What might make it cool for the HN crowd is that every single part of it is written in common lisp. I got a lot of "are you sure" at the beginning because there is this entrenched idea that lisp can't be used for serious work. Happy to report that that view is completely wrong in my case.
I get so excited to build each individual component and so I don't suffer from the burn out or flailing interest that seems to plague a lot of other indie hackers.
Feetr will be a paid service when it launches but I think very justifiable at $8.99 per month. As it hasn't launched, I'm spending time talking through the stocks that it's watching during premarket trading, hoping to show people how it works, what it's looking for and potentially what to expect for the coming day.
Example thread here: https://twitter.com/0xsmcn/status/1555462396696748032
Matano - (https://github.com/matanolabs/matano), an open source security lake platform on AWS. Let's you run security log analytics on an open data lake, and detections as code in your AWS account.
Cool because it's all OSS, serverless, high scale, low cost, and built with modern technologies like Rust and Apache Iceberg.
An alternative Microsoft Teams client. It's cool because it doesn't warm your computer as the official client does. (https://www.opercom.co.uk/) the website needs a complete revamping and the UI is being changed to be more modern.
I'm working on https://liminal.market
It brings the stock market to the blockchain. Like money (USDC, USDT) you bring into the blockchain it needs to be represented by some token, so when you buy stock you get token(AAPL token when buying Apple) with the shares you bought. You own the shares, get dividends and voting rights.
First version is only buy and sell, next is to allow services to use the smart contract, allowing defi services to provide not only crypto tokens on their exchanges but also stocks.
There are currently over 5.600 different companies available for purchase
https://sumi.news – it’s cool because it’s made for skimming, not doom-scrolling. It’s only headlines grouped by day and source.
no website yet, but my project lets you:
- select a data source (e.g. csv file or remote database)
- select a data transform function
- select a data sink (e.csv or remote database)
- execute that pipeline
all of the ui within the browser, would it interest anyone? im curious.
How do you plan to implement the transforms? Are you going to expose the subset of programming language or a drop down menu or something else?
Right now I let the user modify a transform function that accepts the csv row then returns the table row.
Screenshot here: https://ibb.co/09KJX18
Currently it is in JavaScript since it is the easiest to emulate in the browser (I just run the code in a Worker instance, then display the results back to the user).
Kinda ideal for moving from one source to one table, I'm still figuring out how it should work if the user wants to combine data from multiple sources.
What do you think?
I find the JavaScript code not very useful for human input, if the intended audience is not restricted to web developers (a code like interface might still be a good idea). If your intended offering is for devs it would neat some more advanced features to be useful.
More interesting is the question whether you want to support joins. If you do not then you are restricted to concatenation (either in length or in width). Another question is do you want to allow filtering too?
If you want joins i would recommend writing a small query planner because there are some hugely performant optimisations with joins (and filters). If you restrict the possible calculations across loops to simple things like sums, max one could parallelize it between workers with some workm
An interesting avenue for small scale monetisation would be to allow paying users to run them "in the cloud" in JavaScript edge VMs (not sure how they are called exactly cloudflare and others have them). One premium feature could be to pick one of multiple locations where to start them between closest to the data sources or closest data consumer based on the expected ratio of input/output. For large local files and poor connections client side calculations could be used automatically.
https://github.com/hofstadter-io/hof - data modeling & code gen tool, open-source and open-ecosystem. The goal is to code gen the majority (80%) of the mundane code so we only have to work on the interesting parts. You can use any language or technology.