SCUSKU

The aqueduct?

Ah that makes a lot more sense. It seemed so outrageous that they would try to claim ownership of all hosted user generated content, but that makes sense that it was targeted at their community forum, and then misinterpreted.

Very cool! Nice job! Out of curiosity, how much CPU/GPU/RAM resources do you need to allocate to run something like this? Like you mention you're using small models. For example, could this run on a t2.micro on EC2? (1 vCPU + 1GB RAM)

Yes.

TIL about Firefox reader mode! Thank you! Option + Command + R on MacOS, will commit that to memory and use that from now on!

Wow first I’ve heard of this. Insane 2000’s vibes. Honestly though the demonstration social engineering to get a free pizza was pretty cool, and terrifying since that same pattern is applied to pwn sites and stuff.

Will check out the Art of Deception book they recommended, seems good to learn about the low tech ways we can be manipulated to get past security.

Looks pretty great! My biggest issue with Celery has been that the observability is pretty bad. Even if you use Celery Flower, it still just doesn’t give me enough insight when I’m trying to debug some problem in production.

I’m all for just using Postgres in service of the grug brain philosophy.

Will definitely be looking into this, congrats on the launch!

Wow! What a great piece of topical creative writing. Absolutely love it!

I wrote HNResumeToJobs.com using PGVector + the OpenAI Ada embedding model which outputs a vector of size 1536. The new large model now outputs a vector of double the size 3072. I believe PGVector only supports up to 2,000 dimensions for indexing, so that will be a problem.

However, I see that they also support shortening embeddings. OpenAI says that the text-embedding-3-large w/ shortening to 1536 still outperforms the text-embedding-ada-002 model. So maybe I'll go that route first, and then hope that PGVector begins supporting vectors of size >3072.

Let's not forget, "3 billion devices run Java"

Ted Kaczynski, AKA the Unabomber was also the youngest math professor at Berkley. But shortly after he said he hated it and went to live in a shack in Montana.

Somewhat pedantic here, but would the authors use of python to generate C code be considered metaprogramming as opposed to code generation? I thought metaprogramming typically involved writing code that can alter/augment itself, like new syntax.

Personally, I have found a lot of use in code generation, for example using RTKQuery's codegen tool that accepts an OpenAPI schema and outputs a JS client.

Right, I was gunna say this website might be a nice temporary oasis that is too small potatoes for digital marketers and we might actually get a decent signal to noise ratio if it doesn't get too big.

This is a wildly armchair question, but can't we just model human's ability to adapt to change and then as a result determine if technology will be sufficient to save us from impending environmental collapse? I suppose this is a multi-trillion dollar open question...?

To play devil's advocate, had OP not wanted to go home so badly Parkinson's Law would've kicked in and OP may have tried to do things the "right way" which may have taken much longer.

Part of me wants to be able to encode my life2vec and see what the next predicted tokens are. On the other hand, as much as going to a fortune teller sounds fun to me, I worry that if I go I will simply make what they say come true since I will be anxious about what the fortune teller said.

That said, my life is already on a certain track, and because I'm afraid of change, it's unlikely to be much different than it was before. Maybe this is just a weird call to actually do what your gut says, and to follow your dreams.

Wow, that's great. How long did you work there? I worked at my previous employer for 2.5 years. My friend for 2 years (different company).

Yeah this might work for larger tech companies, but when I was working for a Series A startup and got laid off I got 2 weeks. Same thing happened to a friend as well.

Where do you see meetups? I don't think I've ever seen them on HN but would be nice to go to one while I'm in SF!

They are, except instead of saying “AD” for some reason they get a blue checkmark

Another PITA is that even if you have insurance you have to go to in network providers. I’m insured and went to the local CVS to get a flu and COVID shot but they said I was out of network, so insurance wouldn’t cover it. Out of pocket was $63 for Flu and $198 for COVID. I still haven’t got my shots.

The trees are all pretty close to each other, about 3-4 city block radius (map of trees [1]). The issue is that as far as I can tell there is no existing LoraWAN gateway on The Things Network I can tap into [2], which means I will have to setup my own gateway. Which is fine, but it may prove to be difficult to get this onto a tall building since I am just an apartment dweller. At the very least I can put the gateway in my apartment window.

[1] -- https://www.google.com/maps/d/u/0/viewer?mid=1e7K_VdEEYkxuAy... [2] -- https://www.thethingsnetwork.org/community/san-francisco/

That's a great point. I have experience w/ Arduino and have made a carrier board one time [1], but am not familiar with full custom PCB + MCU designs. Do you have any suggestions on where to get started for someone with my medium level of experience here? And also, how difficult would you anticipate a custom MCU to be?

[1] -- https://zachbellay.com/projects/odaf/#iteration-3-phase-2-wi...

So my thought was to utilize the capacitive soil moisture sensor and then coat the sides of the PCB with a waterproofing epoxy as Andreas recommended. My hope being that even if it craps out, I can just swap out the sensor, and since they are so cheap it won't be a big deal.

Also, great to know that there esp-lora board w/ lipo power included!

I think I am largely MCU agnostic, although preferably something where I can use Arduino since I am familiar with that. Good to know that ESP8266 is more energy efficient. I have seen this video which demonstrates that the ESP8266 can run off 2 AA batteries for 425 days [1].

Another thought I was having was to write a mobile app that as volunteers walk by the sensor would send data to their phone, which would then forward it to my backend. I suppose this is a mesh network of sorts? Not sure if this is feasible or a good idea though :/

[1] - https://youtu.be/IYuYTfO6iOs?si=oLuJiGxdQ8VHyu29&t=837

Awesome write up, I liked how accessible the writing is, as well as the fun tone.

I have a problem where I will need to monitor the soil moisture of 20 trees in San Francisco. My current thought is to strap an ESP32, Lora module, battery, and capacitive soil moisture sensor and to send data once every 6 hours or so.

I was wondering if off the cuff you have any suggestions or pointers on how to best approach this. Thanks and once again, thank you for sharing this!

Great movie, never gets old

Ah! Always love to see something from Jeff, glad he's getting the recognition he deserves. He's such a smart guy but very humble. An inspiration to us all!

I have been following the TinyPilot project on & off for a while, and I just wanted to say I appreciate your openness as well as your excellent attitude toward feedback. I think many of us (myself included) have been spoiled with various forms of free/open software and have come to expect things to be cheap or free.

But of course github stars and positive reviews won't pay our bills. So it totally makes sense to charge what you are charging for your product. You deserve it, and you're clearly putting in hard and honest work. If and when I need a KVM I will 100% be reaching for a TinyPilot.

Great post by OP. Particularly sad (and freaked out) that WSJ gets this wrong after having read Manufacturing Consent by Noam Chomsky. (The main premise being the US Gov't manipulates media to spin various narratives to maintain public support).

Of course, Hanlon's Razor (never attribute to malice that which is adequately explained by stupidity) could also explain this discrepancy.