tristenharr

I’ve found the 2.5 pro to be pretty insane at math. Having a lot of fun doing math that normally I wouldn’t be able to touch. I’ve always been good at math, but it’s one of those things where you have to do a LOT of learning to do anything. Being able to breeze through topics I don’t know with the help of AI and a good CAS + sympy and Mathematica verification lets me chew on problems I have no right to be even thinking about considering my mathematical background. (I did minor in math.. but the kinds of problems I’m chewing on are things people spend lifetimes working on. That I can even poke at the edges of them thanks to Gemini is really neat.)

from sympy import sqrt, isprime

T1 = (sqrt(5) - 1) / 4 J1 = (3 - sqrt(5)) / 4

def golden_prime_generator(limit): primes = [] for n in range(2, limit): Fn = ((T1 / J1) * n - (-J1 / T1) * n) / sqrt(5) val = int(Fn.round()) if isprime(val): primes.append(val) return primes

my_primes = golden_prime_generator(1000) print(my_primes)

2930441286939258055400798002200365702070383356835498296052929375099603314612750189999474839964041401824000060018138590135001102587299815016150326470813100791358366141506590207652108187961822226141844888227246720852693565292520772736233258137883131633884772691595926733322344165957105221939274575774318941464860860848233258397675536343685662420797069118608961212878580658178971916344998597383060635840095569469168798433682688690011514459699339511226211249166351120903825586706980141114247391189112452169301152785673208885984700944382085177499062708435697669359679050984522704457968238198039501357594162223897683190805618618838233503289429282246135842585430668875653560981464799218782095773523505176978173330569112583486848660662068908749844158638271675670908226354975956636181462553977591227553535134946781776535903879669168739541718125413163115489438127173145671369147353184448182943977584875431449095152717023151009740993170461840098542328540055499213188472025194001595310097403036158566599062407003698164302276578800844789715519909977450325416955326392430087122738870988283653737701175138504893394815816782040327194725855833

What happens now? :/

Hey! Yes, to be fair I am indeed bipolar so I have personally been a bit concerned about things. Sometimes I do end up going “off the deep end” a bit. Maybe that is what is happening, but I feel I did find some things that are quite interesting though. Also ChatGPT and the like get very overly excited. I didn’t pay much attention to its delusions of grandeur. (Of course forget the things about BSD, Riemann hypothesis and such.. my current goal is to solve ANY open problem in mathematics)

I understand that many of these identities, claims, etc have been proven before.

Personally I am most interested in further developing this Algebra.

The identity e^{2\pi i \phi} = e^{2\pi i(\phi + 1)} though obvious to a number theorist what derived structurally from reciprocal identities involving T and J.

The other curiosity is preservation under transcendental maps.

As far as I understand.. this to me is the only known case where a nonlinear algebraic relation like a - b = kab is realized inside a number field, preserved exactly under exponentiation and sine.

But as you mention I am probably just crazy to be honest haha. Thank you much for taking the time to look and give your thoughts.

Hello! I would love for someone from the mathematical community to take a look at this. Curious what people think. :)

Hi HN, are there any other tigers out there? I’m curious if there are any alumni of Mizzou, or current students here in HackerNews.

Would be cool to create embeddings for historical HN posts, and then use a users favorite posts to personalize the post selection by averaging the embedding vectors for a users favorite posts then doing a cosine similarity search to select stories most likely to be of interest to a user.

Although it would be even better to use a users like history, but I’m not sure if/how those can be accessed.

Speaking of, I’m curious how other folks use embeddings. I know you can average multiple embeddings together, but is anyone else doing other translations and having success? Thinking of King - Man + Women = Queen, It seems a lot of the time I see questions being directly used as inputs for semantic search/RAG. I wonder if it might make sense to create a large set of question-answer pairs and embed them and then determine the average translation to move from “question space” to “answer space”, then when you embed questions you apply the translation on the embedding to move it into “answer space” before performing RAG, or maybe this would just add too much noise?

HN hates my links. Why does this happen?! https://github.com/hasura/hasura-discord-docs-bot

Hi! 100% Hasura is what you need! This is our calling card, I work for Hasura and if you need instant CRUD on your database, new or existing you can’t go wrong with Hasura.

It’s kind of like PostGrest with GraphQL.

It does exactly what you want, you have a database with models and it gives you endpoints for CRUD in GraphQL and you can RESTify any GraphQL to make it into a REST endpoint as well.

It does it by introspecting the database then generating a schema, and the OSS GraphQL-engine that runs Hasura compiles the GraphQL directly to SQL.

Hi! If you’ve ever thought about something like using GraphQL for something like this.. You might like Hasura. (Obligatory I work for Hasura)

We’ve got an OpenAPI import and you can setup cron-jobs or one-off jobs and do things like load in headers from the environment variables to pass through. There isn’t currently an easy journey for chaining multiple calls together without writing any code at all, but you can wrap pretty much any API endpoint via OpenAPI import or a custom action, and you can even make minor edits to things like the API contract format to change aliases/naming.

Our goal is to join all the things, databases and API’s. Most people know us for instant GraphQL API’s that give you CRUD on your database, but we also wrap APIs.

Not sure if something like this would fit your use-case and do check out some of the other things mentioned, but depending what you are trying to do I think Hasura might potentially work.

You can find out more here: https://hasura.io

I’m bipolar and when I was first diagnosed with bipolar disorder I was 19 and in my first ever manic episode I dropped out of college and went deep down this particular rabbit-hole in an attempt to solve the Riemann Hypothesis to win the prize. I spent 3 months on it gradually driving myself crazier and crazier. My second company had collapsed out from under me when a sociopathic employee with a gambling addiction had conned me into thinking they were my best friend and meanwhile skimmed all of our profits so it looked like we were barely breaking even right under my nose. (To be fair I was 19) That in combination with a semester of Intro to Philosophy where I read Aristotle and Socrates and a Algebra class I became obsessed with led to me stumbling into the Riemann Hypothesis, and going deep down a rabbit hole in what ultimately turned into my first ever and most intense manic episode and led to my bipolar diagnosis.

I hadn’t even finished college Algebra at the time. I basically spent months playing around with equations on Wolfram Alpha.

The funny thing is this ultimately was kind of a life-changing moment for me, as I learned programming through it which in combination with a family member dying is what finally pulled me out of the mania, and then I changed my major to Computer Science re-enrolled in college and have since actually gone on to get a degree in CS and a minor in math, sometimes I look back at the giant set of Google Docs I created during that time, mostly in amusement but to this day some things I look at and don’t even understand if it is meaningless or not. A part of me wants to pick it back up, but then I tell myself it isn’t worth the risk, I’ve always felt I am uniquely qualified to understand how the guy in the movie “A beautiful mind” felt when he stopped talking/engaging with the hallucinations.

Here’s some links to random Google docs from my ~2017 rabbit hole if anyone is interested. There’s a lot of fairly interesting usage of the Euler-Mascheroni constant and honestly it goes a bit all over the place.

I kind of doubt there’s anything here, but if you’ve ever wanted to peek into unchecked mania.. have fun. It was months of insanity. (Sharing these in good faith!)

https://docs.google.com/document/d/1CERdyZMlGj76EmMJXt5Zr3pM...

https://docs.google.com/document/d/1LpKqgCzjo8LrxB547WY5YdVP...

https://docs.google.com/document/d/1JvV5NJTYaTHn3eSAShcUtE8J...

https://docs.google.com/document/d/1Ej0bNIpYNT30M31pCxx-LDoe...

https://docs.google.com/document/d/13nlw3vbxyRGd6pYeAre8c7e5...

https://docs.google.com/document/d/1WChBcGtaUdLckPDqNBVGpP9x...

https://docs.google.com/document/d/1_bQbgJ5HIyv_D5ex9TWK2XxH...

https://docs.google.com/document/d/1AomtTAP4GBXpOuArvBYQsHiP...

https://docs.google.com/document/d/1QO8iQ2OmPZkQ7lNS0uUB0uyy...

https://docs.google.com/document/d/1zzedxz1Ya9rT8N4RZZNtpsrr...

https://docs.google.com/document/d/1vB0Qbqe0yiO0tcxswl2jniYc...

https://docs.google.com/document/d/1K4upBz2awHnKlr5JE1xocntp...

https://docs.google.com/document/d/1XhOIy0RO59CrZEXXPK-QAR_s...

https://docs.google.com/document/d/1rJx2KHQ0GqZfESqyFzXGLmN_...

https://docs.google.com/document/d/19kZrg8HvxZ31-goiwL0XLyPC...

https://docs.google.com/document/d/1F8c_gVy47bRw5zZVWKb7x5SV...

https://docs.google.com/document/d/16JolkQcHD3eviZuakfT7iam7...

Clearly lots of people have tried to replace YAML with something else and that hasn’t worked. What’s your wishlist on making YAML actually work for declarative systems? Or can it work?

Would things like..

A great LSP with semantic autocomplete, Native cross-file imports, and conditionals based on environment make things feel different/better?

What should modern declarative systems be doing in your opinion?

I was surprised when I learned Flutter actually runs as native code without needing a JavaScript bridge like so many others.

Although similar to something like React Native - Flutter does paint the screen. Personally I see it as a potential Electron replacement in the future.

From what I’ve seen Impeller is also a big help on iOS. Screen jank was pretty bad for us when Flutter was using Skia as a rendering engine, although when I last messed with Flutter (~6 months ago) Impeller felt almost production ready but iirc there were still a couple small things, not sure if they are better now.

Personally I’ve become a big fan of Flutter, especially for back-office/utility type apps. We would follow all the Material Design 3 guidelines and when it comes to time spent, if you need to develop the same app on more than 1 platform, it’s worth consideration.

I’ve found some people will say: “But it’s hard to get it to look like a native app in Flutter” but for me I’ve always seen that as more of a skill issue. You are literally painting the screen, you can get whatever pixel perfect design you want, if it can be Figma’d it can also be Fluttered and to be fair there are Cupertino widgets and you absolutely can get the best of both/many worlds if you want. Just like how on the web you can do a transform based on device width to switch between the mobile or web view, if you desire you can have different widgets show up for different platforms and give everybody a native feel. I personally still think that’s easier than writing/maintaining multiple individual apps.

I think the biggest downside though is that search engines don’t do well with Flutter Web. It isn’t the nice easy to crawl HTML, and it’s really hard to get a flutter app indexed or to give it good SEO. There’s also silly rules that search engines have about how the content displayed to the user vs crawlers has to have the same data. (I.e. when the web-crawler asks for your page, you are supposed to give it a page with the same data you would give the user.) The best we could do to improve SEO was to make the landing page a page that in the background ran all the same queries and shoved the data into HTML tables, with a popover that said “this is a page for robots that had a button that said “Take me to ABC.com” and a check box that said “automatically redirect me next time” that we stored in a cookie. It was hacky and not a good user experience.

For now, whenever I need to do anything GUI I will use React for the web and Flutter for everything else, but it would be nice to truly be able to only use 1 framework for every platform, which to be fair I’ve used Flutter web and it isn’t all to bad if it weren’t for the issue with search-engines web-crawlers. For apps that don’t need SEO.. Flutter is a great choice. I suspect we’ll see a lot more SaaS apps building in Flutter in the next few years. I think it’s on the cusp of being mature enough to be seriously considered for production scenarios.

I’m just mostly worried about the ecosystem. Google chose Dart for flutter after the results of a lot of research that only Google scale companies can do and to be honest although it IS a lesser known language it isn’t a bad one. It’s AOT and JIT so the hot reload is amazing. It reads like “insert most language you know” here. If you know how to code nothing in Dart should be drastic or surprising. If anything I’d almost say Dart is boring, which I like boring things. I’ve experience with a handful of languages, and for me Dart ranks highly in terms of usability/ergonomics. But there’s a not unsubstantial risk that Google could decide to drop the ball and Flutter could become kind of like the Windows phone. Really great, but canned because people didn’t build enough things for it.

If I were on the Flutter team I’d be most focused on figuring out how to deal with SEO issues. (assuming impeller does indeed provide a full fix for the screen jank, I’d assume it’s gotten better in the past few months, can anyone comment as to if it’s “there” yet or not?)

In middle school one of my science teachers was really great and thorough, the kind of teacher that you remember the things you learned years later because they made the classes memorable.

We would drop things off the roof of the building and record with slow-mo cameras to calculate the formula for gravity, he lit hydrogen balloons on fire and you could feel the heat. The coolest day of the year was when we got to play with his “hovercraft”. He had repurposed a Vacuum cleaner to create a baby hovercraft.

It was basically a small platform, with a plastic tarp on the bottom that had little holes and when you turned the vacuum on it would expel the air and the little platform would ever so slightly lift up and glide across the ground.

The reason I remember it so vividly was because each of us students got a chance to sit on the chair atop the hovercraft platform and the teacher would push us down the hall and we’d “hover” from one side of the building to the other.

I am forever grateful for the teachers who go above and beyond to make learning fun for students with no reward to themselves.

The coolest thing to me was watching it do laundry in the video it takes a jacket and puts it on a hanger and then zips the jacket. Watching it get the zipper was pretty satisfying.

It’s also amazing that there’s just instructions and it’s open to anyone. Kudos to the builders. They sauté shrimp, do laundry, and wash dishes with this robot you can build at home.

If that isn’t cool I don’t know what is.

Also, sidenote although pg_turso is experimental I added text, varchar, uuid, timestamp, timestamptz, json and jsonb to the supported types ~1 month ago for anyone who saw it in the past but that was holding them back, don’t let it be what stops you from hacking on it.

I tested it on the chinook music db, I was able to fully replicate a table, although I admit it’s barely functional as a cache because it doesn’t handle relationships or primary/secondary keys and I don’t think it does anything with unique constraints either, but with that said it is functional. It will take a table from Postgres and put it into every region supported by Fly.io and with embedded replicas you can just push it through onto the users device if you want. I jumped out of my chair and was laughing like an idiot the first time I got it to stream the entire chinook music DB from a Postgres creation script straight into SQLite all over the world, all while running postgresql on localhost inside Docker. I can taste the future. XD

This would be amazing. Anybody who plugs into the PostgreSQL replication protocol is on my watchlist. I’ve been playing with https://github.com/turso-extended/pg_turso which lets you replicate PostgreSQL tables into SQLite on fly.io using Turso.

I’m still predicting we’re getting a Kafka clone in Rust at some point. No idea who or where from but I’m predicting/hoping.

The world is moving to events. I’d be surprised if 2024 was “the year of events” but I’m sure it’ll be that year sometime in the next 5 years or so.

I want to speak WAL everywhere I go. Let me talk WAL to my vector database, and my OLAP, I want to talk WAL to NoSQL, so what if the schema isn’t enforced, I still want a logical representation of it.

The year of local first.

Postgres + SQLite will get married.

Somebody will start a project to rebuild Kafka in Rust, hopefully open-sourced.

People will over-do it putting things on the edge, freak out about the pricing, then scale back, and there will be backlash because of poor implementation putting too much data on edge. This will happen to companies who try to get in on it without understanding why they want/need it.

The idea of pseudo-elimination of loading on the web via using a state machine and pre-caching every state transition within X taps locally will become more popular/known.

There will be an AI model trained specifically on propositional logic that’ll surpass the abilities of any other current models in reasoning and mathematics.

Somebody will come up with a law about the speed of AI’s development and correlate it with Moore’s law.

I’ll be collecting and cataloguing a bunch of data and curating question-answer pairs over technical documentation, experimenting with GPT-4 for generating QA pairs. A focus on building datasets to represent multiple experts.

Lots of data curation.

This is exciting, I’m hopeful that we get distributed SQLite via something like Turso that can hook into this to stream subsets of data into SQLite edge-caches that can be pushed through the edge onto devices via something like embedded replicas. I know experimental work has been done here with pg_turso (I recently wrote my first few lines of zig to add support for some additional data-types for it) and this bodes well with the recent announcement of Postgres coming to fly.io, I dream of a day when I can make Postgres and SQLite play nicely together.

Ah, Fun that links didn’t paste properly, is there no way to delete posts?

I’m a big fan of Qdrant, I also have heard rumors than OpenAI uses Qdrant as their vector database of choice.

I’ve been building a Hasura Data Connector for Qdrant and it’s been too much fun. Glad to see them getting talked about here.

I think there is a lot of opportunity in this space. It’s rare to see this many comments speaking of the pain points here.

I recently was a co-founder of a delivery app in the CPG space, and I think this space is ripe for disruption. We were running micro-fulfillment centers and staffing delivery drivers, but I wanted to pivot into a SaaS model taking on the data-plumbing between manufacturers and retailers. It absolutely amazed me that NOBODY seemed to have even half-decent API’s.

I wish people could all just get together and decide on things like a universal barcode system. You’d think things like electronic inventory transfers would be the norm, yet we were still dealing mostly in paper receipts.

At Hasura we’re in the midst of rebuilding our engine in Rust and I just have to say I am super impressed by the number of libraries that have no problem shipping to wasm32-unknown-unknown.

A personal project of mine has been trying to hack apart our newly built engine to shove it in the browser so I can mediate local state with it. It’s going surprisingly well, the only dependencies that gave me any trouble were Axum and open-telemetry both of which aren’t relevant in this case. I am so excited to get SQLite + Hasura engine running in the browser. (Can’t wait until we OSS the v3-engine which is coming soon!)

I’m pretty well sold on Rust, I like the trait system, borrowing isn’t nearly as scary as I think it gets made out to be.

Curious who else is in the midst of a rewrite? It seems like it’s rewrite in Rust season for lots of people.

Anything you’ve learned in the process? Has it been a good decision?

I love getting approached by friends and family and having to do my best to gently let them down on their “million dollar ideas” they “need my help building”.

It’s not that they are all bad ideas. I just stopped putting any value at all in “ideas” a while ago. Ideas are a dime a dozen, even trillion dollar ideas. (Here’s an idea, lets invent teleportation, it took me ~3 seconds to come up with a trillion dollar idea!)

I think there are three things that matter so much more than the idea. Team, execution, and traction.

You need the right team, and putting together the right team is the most difficult thing to do, because a good startup team needs to be able to cover the largest amount of surface area in terms of “stuff we can do ourselves”. There’s a lot of different “departments” in a company, and at the early stage it’s the founders covering all of them. If you want me to play Frontend Dev, DBA, Product Manager, Architect, Backend Dev, Tech support, and Tech recruiter I’m ok with that so long as you can play Accountant, Fundraiser, Spokesperson, Secretary, Lawyer, and customer support.

You also need execution. I won’t say the execution has to be great across the board, but ideally it’ll be great in a few core areas key to your idea (I.e. your special sauce), while being “good enough” across all the “departments” you juggle to convince people you know what your doing. This is where I think lots of people fail. People have this false notion that they can do the things they are most comfortable/good at and somebody else will soon come along to handle the other stuff. (Well we can just hire somebody for that they say — with what money I ask?) Execution means understanding that no job is below you because at the end of the day it is solely on you if it gets done or not and oftentimes there is nobody else to do it. The monthly update emails don’t get sent unless you send em. The employees don’t get paid unless you pay em. (They tend to get bitchy about that if you forget to do it on time too.) The product doesn’t get built unless you build it, and the customers don’t get acquired unless you acquire them.

That brings me to the last thing, which is traction. If I had a dollar for every-time someone has tried to get me to go in on building their “New take on social media, it’ll be the next big thing!” I’d probably be able to self-fund my next venture by now. If you have “the best thing ever” but nobody’s using it, it isn’t “the best thing ever”. It’s kind of like Schrödinger's cat. Your product is both the best thing ever, and a piece of garbage until somebody opens the box and uses it, upon which you find out which it is. You don’t need traction at first but eventually you need traction or your venture will become a exercise in masochism as you slowly slide towards bankruptcy trying to get enough traction to save you the company from entering a death spiral.

Hello! I’m the author of this blog, and I recently joined Hasura as a Technical Evangelist. This is one of my first blog posts and I’m new to this kind of thing so any feedback is welcome, good or bad!

Nice job DuckDB team, those are great performance improvements compared to a couple years ago. It’s neat that solutions like this are being offered and developed. At Hasura we’ve been working on a data-connector that’ll wrap around DuckDB. Curious what others would think about having a GraphQL layer on JSON/Parquet/CSV via DuckDB?