Is this going to nuke all bring your own API 3rd party tools? I've been casually using fewshell https://github.com/few-sh/fewshell with my Claude api key, I really hope it's going to keep working. I've just finally managed to turn myself into a reasonable devops team with it.
Oh thank you! I'm using these tools but occasionally I feel like a medieval horse rider trying to drive a sedan. Glad to know, I haven't used OpenClaw, I prefer the meat computer for autonomous compute.
I send things too iNaturalist all the time, it's great, it really helped me learn about my local fauna. I want to do a project with their API to identify a couple hundred wildflower photos I've been hoarding. Would that work? ( Idea is my wildflower app could send to their models to confirm my original identification)
Hey, good news! Pollination ecologist + ML guy here; with open models coming soon.
You can keep an eye on (gh) polli-labs/linnaeus (a bit stale; I'll rebase on my private repo later tonight-).
There's some cool ideas in here to exploit the structure of taxonomic hierarchies to help the model approach recognition how a professional taxonomist might.. so working from coarse to fine, taxonomy-guided label smoothing (distributing alpha mass by taxonomic distance)..and (forthcoming) RL on expert consensus to teach abstention (if an expert could only identify a specimen to genus for some set of inputs; then our model should abstain from a species classification for the same inputs).
Unfortunately I am very, very compute-constrained- but shooting for late April/first week of May for insect + flowering plant models. (Other taxa will come later; probably as unified model).
I'm working on camera-based (automated) ecological monitoring systems for ~6yrs at this point; it's a really fun problem space! dropped out of grad school to go all-in on automating my favorite job I ever had (pollination ecology field research..watching flowers for visitations!); since I knew I'd always be a mediocre ecologist- but an engineer that happens to care about ecology could be very very valuable to my field.
a taxa recognition model turns out to be only a small piece of the system you need to extract structured observational data from cameras in the field :-)
Working with one of my partners right now to launch a really cool demo of what's possible these days- Texas folks especially; keep an eye out on wildflower.org around May 1!
I'll spill more ink soon but (anyone) please get in touch if you find these things interesting. Or if you'd like to help me out with compute/expenses!
This project looks awesome pollination ecologist + ML guy! I starred it and will definitely be exploring using this. I'm actually pretty simple with my needs because I'm after only a few specific wildflowers not the whole ecosystem. But of course it's impressive how comprehensive your project is looking.
Are their models considered to be the best or is there some competition? For plant identification, they blow every other free app I have tried out of the water. It also seems to return the genus of a plant rather than misidentify the species which I find impressive.
That's too bad, maybe I can upload it to iNaturalist then reference the entry there. I don't mind if it's duplicated, I just want to be able to improve the location data without sharing the improved location data so publicly.
Yet they shelter under a 'Science' tax-break. It's duplicitous. They should publish their models and build process. If it's not available for replication, it's not science.
I have been experimenting with eating ingredients from time to time to help my body learn to be specific in what it's asking for. As a side effect I now crave almonds much more, I think because I finally exposed my body to a clear source of Vitamin E.
This story is very focused on digital architecture, particularly risc-V. There is another big issue with open source processes, sky water 130nm is the most accessible, but expanding PDK (process development kit) is also a big deal for open source users to be able to build hardware.
Check your gas cap, I had this effect happen in my car because I had lost my gas cap. It can also happen in its damaged in some way. I don't have this habit so I can't tell you if I am observing a similar effect.
No not at all. I only use AI assistants for help with price comparison of things when I'm in the grocery store and want to know what of the salsas is the best price without preservatives or other things like that.
ooh, does that help? or does it hallucinate a lot?
I made this for myself a while ago: https://rewardsgenie.calstudio.com/
it's a tool that tells me what's the most rewarding credit card to use for any purchase, but even with good context, it would sometimes make stuff up so I kinda stopped caring
It's pretty low risk and I'm just doing it for saving a few bucks. So far it has only had one moderate oppsie that I caught. It recommended a kefir brand that it turned out was actually low fat (which I didn't want), but in defense of the AI, I had also read the entire package and didn't see the tiny low fat words written on top of the pink cap either. I only noticed when I started drinking it and went back for a closer look.
I tried to checkout your tool, it starts with an account though and I don't have many CC. It sounds like it sound be useful though if you have those stacked deals like 3% cash back but 8% cash back on Tuesdays for gas only and then another card that's 5% cash back every day. I agree it's a good amount of context to keep track of though.
I kinda turned the tool off for now anyway cause I stopped paying the subscription for it haha
I imagine this to be better with an apple wallet type interface where I don't even have to ask and the wallet just auto selects whatever's best - but this seems to be super hard to build so I kinda just gave up
I have the same experience with Claude, but it's been rare for me to hit limits. I tend to ask it one or two improvements to the app and then I have a time burden of using the app to make sure it's still what I intended. I have the $200 per year plan which I think is the same as $20 per month just I paid in bulk for a discount.
That's a great idea! I'll try it out, it might save me a lot of time. I haven't really done user testing before, how do you test for things like 'still feels responsive ' and 'is easy to use one handed'? I'm currently running with 0 programmed tests but I do spend like 4+ hours just using the app between Claude sessions.
I write my goals. Claude drafts a plan. I then feed it to Claude Chrome extension for live testing. Claude Chrome extension will set up the console inspector, measure response times and change UA if it's part of the plan. I'm sure you could tell Claude Chrome extension "ok make this change on the fly" by injecting JS into the console and do AB testing and output the result in an easy to understand report.
In other big city you work in Finance tech or Industrial tech, and so on. The tech is always a servant of some other mission.
In SF and California more broadly everything else is subservient to tech. You work in Technology film or Technology agriculture or even the prized companies of Tech for Technology. That's why there are more billboards focused on technology here, it's because there is less reason to advertise a car, the big money is in moving technologies.
reply