dchuk

Just tried both the Omni model and the text only model, both as mlx versions, in lm studio omlx and rapid mlx, none of them could actually fire the model up, kept saying safetensors cant be read (yes I realize Claude should hear this not this thread).

You should talk to a lawyer. And not post something like this publicly

Electric semi tractors are impractical in all but drayage use cases (port to warehouse short haul). Even then, they require specialized yards and hours to charge, and they are twice as expensive as diesel tractors. Refueling a diesel truck takes minutes. Plus the electrical grid impact to supporting the charging of these trucks, keeping just 50 trucks charged annually uses as much electricity as a decent sized town.

For autonomous trucking, you’re grossly underestimating 1) how dangerous these trucks are. A small rear end accident with a car does some damage, a small rear end from a tractor with 80k pounds of cargo kills people. 2) They are hard to drive, hard to navigate in cities, require inspections constantly for roadworthiness and safety. Sure having a tractor pull a load on a few hundred miles of straight empty highway in Texas is doable today, having that same truck get to a warehouse in the center of Dallas is a nightmare even for pro drivers. 3) Plus no one knows how to legislate or regulate these things. If a robot tractor kills someone, who is at fault? The vehicle maker? The autonomy software maker? The carrier? There are already nuclear verdict lawsuits for these types of situations with human truck drivers that can be tens to hundreds of millions of dollars I.e. shut down a carrier bad. You think a bunch of octogenarians in congress know what the hell to do with any of this stuff?

I say all of this while also fully agreeing that consumer autonomy seems to have been solved or is on its way to being solved, I’d bet in 5 years it will be weird to get an uber and there’s a person actually driving. Commercial heavy duty trucking is a whole different class of problems that goes well beyond “can the software do it”.

Source: I make software for trucking carriers and drivers

Not the same but adjacent hobby, I’ve had good success with feeding 3mf files to Claude (usually just fable in desktop app) that I found on maker world to prompt for modifications to the model and it works damn well. Stuff like changing a functional model from press fit only to using inset m bolts and nuts. Way better than doing it by hand in a cad app

Funny, I was working with Claude and Astra today ideating on how Loco and topcoat could be integrated for the ultimate modern rails alternative stack

Not to mention all of the SOTA AI subscriptions handle this perfectly now when there aren’t csv exports

I’m playing with this exact concept but going down the path of a Claude code plugin that is optimized for configuring lowdefy apps because of that framework’s unique approach where you don’t write code you generate yaml configs that then drive the rendering of an app. So far it’s working pretty well, still pressure testing it. The audit control and permissions management you have is great, especially around the connectors. Nice job!

I just redid my homelab/media server (switched from an old NUC I couldn’t figure out how to stop overheating even with some decent modification work, now using an hp elitedesk with an i5 processor that is handling my stack nicely). Thinking about setting up a vm on the base ubuntu install for isolation to run Claude in. May play with dispatch, may just put raw Claude code then use Moshi app on my iPhone and iPad.

I use this little three liner regularly with people in regards to AI (applicable as much to coding as it is business work):

- I don't care how you produce the words

- I expect you to use tools to be as efficient and productive as possible

- You are accountable for every word you produce

Oh how I miss old school forums. It’s crazy to me how communities are wholesale embracing discord, which just is not the right form factor at all for anything but ephemeral real time chat. I remember engaging in threads on real forums for literally years. It was so great

very interesting ideas, thanks for the detail. I'll think about the wordpress search plugin idea. here's my twitter handle if you want to connect and chat there: https://x.com/iamdchuk

Some other ideas I've thought about:

- Pay to boost a specific article for a period of time (with appropriate UI changes so people know it's a boosted article)

- A job board that puts roles in the article rankings (ideally also semantically matched to users based on their profile/topics they follow). a reverse job board could work too, where people can indicate the types of roles they're open to

- An Offers board, for people to pay to advertise their product / startup with a specialized deal available to readers of the site (could also be more of an affiliate system to reduce friction of getting Offers submitted by taking a commission on payments collected using the offer code).

Overall there's a bunch of directions this could go, I just need to start promoting the site itself to get traffic up and consistent.

I’ve been noodling on an idea sort of similar, which is to have a system where if you register for an account and very your site (with a technical mechanism in place to enforce that), you get a slight boost in the ranking algorithm for your content. Then, if you pay a small monthly fee, you get an additional boost in rankings, as well as some amount of (anonymized appropriately) analytics about your content performance on the site. Not sure if anyone would ever want to use that and I don’t have anywhere near the traffic for that to be worth it yet, but seems worth playing with.

I also want to add content digests/newsletters that highlights best content of the week for sharing on twitter and via email to those who register accounts. Typical growth marketing stuff

It’s funny, sites like these are so divisive. I’ve gotten as many “do not llm summarize anything for me” as I’ve gotten “this is super handy!” responses. I like them, so I’m keeping them.

Interesting idea overall. Could be generalized to just general trend identification I think based on named entity recognition (like google trends). Over time you can trace first mentions of stuff and buzz over time. I’ll noodle on it.

Nice job, clean and well built. I built something very similar, with a focus on engineering/product/startup blogs https://engineered.at

What's your tech stack? Mine is pretty vanilla ruby on rails, while my site is closed source, I did open source a rails engine for adaptive monitoring of RSS feeds that does a lot of the heavy lifting: https://github.com/dchuk/source_monitor

Generally it’s things like that yes, but also stuff like preparing the pulled content to be summarized and then taking those summary batches and preparing for synthesis in a templated format

Ha funny, I was speccing out an idea for real time Claude code interaction from local apps using some tricks vs using the agent sdk when I got the popup to try Fable. So of course I gave it a go, and it triggered the sensitive content warning immediately, which I was very confused by until I put two and two together.

Fun times when “safety” means both the safety of mankind, and also the safety of revenues

I’ve been using this general pattern - a custom cli app for deterministic tasks, skills for the agent harness, run the skills in the agent and it produces artifacts for you by using the cli and its own agentic reasoning - a lot lately for work. Things like “give me an executive brief of the activity in these teams backlogs over the last month” and in 5-10 minutes I have a few page doc I can read that is cited with the tickets it analyzed and I don’t have to go bug people or ask them to do yet another task for me, just make sure your backlog is updated and detailed like normal practice. It’s awesome and really fits a useful spot between pure agent usage (which is hard to get consistent results from on repeat tasks) and not having to build/buy a full blown app for every random thing.

I backed jetkvm on kickstarter or indiegogo or whatever site they launched on out of excitement. It’s a well made device, the software is clean. It feels like it’s stopped being iterated on though which is a bummer (the tariff timing was likely brutal).

Even though I have it on Ethernet on a gigabit network at home, with WiFi 7 mesh that gets 950mb up/down to the internet, I’ve never been able to use jetkvm whether through their cloud portal or through direct ip connection without it feeling very sluggish. It’s plugged into an aging NUC that I know is still performant, and I previously had no machine setup with settings configured to a smooth basically realtime feel even over Tailscale remotely, so I don’t know if I just am using the wrong settings or something or what. Might just have to have codex computer use fiddle with it to see if it can figure it out.

Been using his skills a lot lately, they are wonderful. I’ve added an issue to specs skill that grounds the issues with a technical plan against the current codebase, and a research school that spawns a bunch of agents to look up best practices on the internet for those issues with specs, it really dials things in. I need to issue a PR to his project for those two…

Interesting. in terms of "crawling", the way the engine I built works is by default it's just polling the rss feed of a site on an adjusting cadence like any other rss feed reader. On some sites, the engine can do a follow up scrape of the article link from the rss feed if the full content of the article isn't provided in the rss feed. So it's not real crawling, more fetching/scraping if necessary.

But I hear you.

ok this should now be fixed!

Figured it out, had a random block of Firefox versions less than 147 in my ApplicationController for some reason. Of course my home internet went down though so I’ll push in a few.

I presume you’re politely asking in order to block? Which is fine, I get it. On my phone right now but can update later.

I do want to ask though (and I should make this clear in a FAQ or something): the way I check RSS feeds uses adaptive scheduling, so I intentionally don’t check feeds of sites too rapidly. Then the summarization is based on the full article content but I never render that full content on the site (to avoid traffic hijacking concerns). Given that: what’s the concern?

Thanks for this info! Very helpful

Yeah I know and agree, just wondering if something is haywire in that logic somehow. Otherwise it’s a bizarre issue but I’ll get it fixed

Great feedback, should be straightforward to make happen. I’ll try to implement tonight.

you know, I had setup some analytics filtering based on geoip because I was getting crazy spam traffic from Chine and Singapore, but that should only be affecting analytics not the whole site. Mind if I ask where you're located? (you can email me privately if preferred: me@dchuk.com)

That’s…bizarre. Let me take a look

EDIT: just checked in firefox, I don't see an issue. can you email me at me@dchuk.com and maybe I can debug with you?

I built a site that's similar in concept to Hacker News, but is entirely fed by RSS feed content, that is then bullet-pointed summarized on the article page: https://engineered.at/

But I also extract topics automatically from the content too with LLMs, to allow for dynamic topic pages that users can separately subscribe to to tune their feeds.

Haven't promoted it much, but it's pretty amazing what you can do for a couple bucks a month. And my main thesis with this site is that by locking the content to only rss feeds of known blogs, you dramatically reduce the spam submission risk (basically eliminate it). Doesn't handle the spam comment side of things, but that's a different problem.

EDIT: I also open sourced a Rails engine I made to power this site if anyone is interested: https://github.com/dchuk/source_monitor

This is awesome thank you for building this! I’ve wanted to try building some sort of companion robot for my kids to play with, that leverages edge voice AI models and likely an api connection to one of the big SOTA providers for the brain, with a custom system prompt to have a very simple child-like personality. I feel like it would be neat to use an android phone as the “face” where you get a nice screen to render on, compute for the edge models, and the forward and rear facing cams for the vision all included. One of those daydream projects…