July Hata

Excited to announce Roc Camera is open for batch 2 orders! Current lead times are about 2-3 weeks depending on where you are Check out our new website in the comments :)

This is intuitively obvious but bares repeating mostly to myself: I want to pick futures that arrive the fastest on the long run not in the short run

Different cities have different textures

Part of the reason I'd argue this is incredibly hard is two fold: 1) Taste / craft is about cultivating sensitivity to the world. The world of ambition often leads to optimization - and optimization often means curtailing the sensitivity to the world itself. Making sure you fight tooth and nail to ensure space for that craft to grow is much easier said than done. Also creativity, and making new things - I mean this in a more creative endeavor environment requires nous and balls. Hard to have conviction - craft and taste seem to be downstream of that 2) The other part is environmental valuation of risk vs reward. I believe that environment is destiny. If you are in a place where the reward / your peer acceptance cycle says one thing (we value creativity!) but rewards another (shorter timelines for outcomes, legibility, lack of larger change etc) you end up optimizing for the meta-game of what is going to return more immediate results in the world. So you end up focus on that That being said - I believe in people high craft / high ambition - and I also believe it can be done, but hard to cultivate. Something that people say that they often want to do - but most do not want to do this. You want the rights to fruits of your labor, but not many want this path

I think about this still from time to time

I want to make things that are tactile but evokes memory. What does that mean? What I mean by that is that you can touch it and you can feel it and remember touching it somewhere at sometime. What we can touch, what we do, for me it often evokes= memories. We compare them to memories. Your favorite toy that you used to bring to bed as a touch and how soft it was. The feeling of a t-shirt the first time you wore it after you bought it. Shoes that fit just right, because you've used them for a while. Tactileness - is a sort of memory. When I touch objects, I love objects that remind me of a feeling of touching, it transports me through time to places and memories of long ago. Sometimes memories of the future, like this works in a place that doesn't even feel that out of place, but its never been there. This feeling of remembering, both of the past or the future gives the object a sort of 'rightness' in its existence. Almost like a justification of its existence that goes beyond simple utility, which of course it should serve utility too (nothing wrong with art tho) - but that feeling underneath should hum like a gentle ocean crashing over and over again in the distance.

@mynamebedan The next crazy part is realizing if you become Japanese you have to do that for other people so it feels this way

they were smoothing over the edges as you broke every rule there is to break without even realizing it. it feels much easier to be in japan when you’re clueless. once you understand everything then you feel a pressure to know and follow every unwritten rule

japan is interesting because the first time you come here as a clueless tourist you’re like “wow everything is so smooth and seamless” and then you realize after spending far more time in the country that actually everybody else was making it feel that way for you

@jtriley2p @roc_camera Honestly same

@0xJuly @roc_camera i would simply trust my peers over trusting hardware i do not understand

app's closed sauce + ios only :( sdk's oss + has hole :3 steps: --- 1. img captured 2. img written to disk 3. c2pa manifest gen'd 4. img + c2pa signed 5. (optional) send to succinct's prover net hole: --- there's a gap between capturing the img & signing, an impl'r can ai-gen an img, forge c2pa manifest, sign it, & gen an sp1 proof about it other stuff to know: --- apple attests to signer's p256 pub key from secure enclave sp1 proof exposes only img hash, app id, & apple root cert; BUT succinct prover sees device id, sig bytes, raw img bytes, & metadata (sus)

@endlessmeee @TheBlockCo hell yeah- let me know i'm happy to talk more

This news is not some rocket science, a gift I got last year is a zkp camera seems to be 3d printed, which also "combats" AI fakes. The product is made by @0xJuly. The Block @TheBlockCo should interview this person

Wheels up for NASA’s X-59. ✈️ In its first wheels-up flight, the aircraft revealed its sleek, streamlined design—key to reducing sonic booms to a quiet thump. See how this milestone moves us closer to quiet supersonic flight over land: go.nasa.gov/41B3YV6

T.S. Eliot - East Coker from The Four Quartets

Speaking as a taxpaying American, this is (one big way) how I want my dollars enthusiastically spent by my government.

I live for this shit

🥹

@SchezHugo Come by! And we can do both

@0xJuly Surfing and golf, damn 😍

A few mins from my house:

@motocap_ It slaps

@0xJuly sharp park course HITS

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

For example, we gave Claude an impossible programming task. It kept trying and failing; with each attempt, the “desperate” vector activated more strongly. This led it to cheat the task with a hacky solution that passes the tests but violates the spirit of the assignment.

Not choosing to do things in the age of having a magic button (aka slop cannons) that can generate things is going to increasingly be the difference between great design and something that thinks of design as being limited to just hiring a designer