My company gets a lot of incoming chats from customers (and potential customers)

The challenge of this side of the business is 98% of the questions asked over chat are already answered on the very website that person started the chat from. Like it’s all written right there!

So real human chat agents are reduced to copy paste monkeys in most interactions.

But here’s the rub. The people asking the questions fit into one of two groups: not smart or patient enough to read (unfortunate waste of our resources) or they are checking whether our business has real humans and is responsive before they buy.

It’s that latter group for whom we must keep red blooded, educated and service minded humans on the job to respond, and this is where small companies can really kick ass next to behemoths like google who bring in over $1m per employee but still can’t seem to afford a phone line to support your account with them.

So it’s a plug-in for WordPress that rips wordpress.org links out?

That’s hardcore lol. The world really is pissed at Matt Mullenweg.

I think I’m understanding the playbook here.

Eventually YouTube will cost 89.99 a month and also be full of ads.

Then they will be disrupted by some startup company that will have to fight uphill for search and bandwidth access through Google’s monopolistic front door.

Years later a court case will be decided and the new guy will be up and running a free service with a few ads or 1.99 a month for no ads!

And then we get to go through it all again.

It always involves

sudo rm -rf *

Doctors face a similar obstacle before they can practice: medical school and residency. They literally have to jump from zero to hero before the first real paycheck.

Things may evolve this way for senior software developers with a high rate of dropout.

I’m guessing the money flows in this direction because telegram actually has users.

I’ve heard of vibe coding but in the context of being able to identify music that fits a “vibe”. What are you talking about?

This is when you give some LLM a prompt such as “write a game like Minecraft except cooler” and the system will output some code that might run and might vaguely resemble a block game.

So then you go back ask for more, it does something to the code potentially improving or breaking it, go back again ask for more, and repeat over and over. I’m being a little bit sarcastic because most serious developers look down on this, but really this is how a lot of coding is happening these days. There are tools to make this process somewhat usable and they are getting better every day.

Interesting. I can buy that idea, a model that’s designed to be general and answer all questions is going to have to make compromises in a lot of ways.

So it’s possible that model benchmarking needs to be revised in some way to give more useful analysis of its capabilities.

The industry is quickly moving towards using agents, MCP connections (sources of real-time data for the model to pull from, and apis that allow the model to perform tasks, like putting things on a calendar), and RAGs (augmentation with sources of truth, such as a 100 page pdf guide for example), and models that seem to be more aware that they can get data from other sources.

The future might become specialized models all the way down.

Just today I’m playing with “vibe coding” and using one agent as an orchestrator that assigns and monitors tasks to other agents. The result is still slightly bullshit code but it’s amusing to watch it work. Not sure yet if this is a strategy to spend all my money through API fees or will result in something useful 😂

You can do due diligence as a buyer forever but if the seller lies or doesn’t disclose… Problems like these happen. Lawsuits are potentially incoming to figure that one out.

It kind of looks like the new owners of VPN Secure got screwed - the last owner made all these costly lifetime deals and didn’t tell them. The obligation/liability to service those deals wasn’t transferred to the new owners.

Which means the old owner is probably the bad guy here and still owes these customers for their lifetime subscriptions.

Just curious if you’re a developer or using LLMs often.

I like Anthropic’s sonnet 3.7 model for agent and code related tasks more than the Open AI models at the moment.

Deepseek and LLama can be run offline, which is great for certain uses especially the aforementioned BS tasks that can perhaps burn through API tokens. Quality of output doesn’t match the top models but this is second to privacy for many.

Not sure where things are at with Dall-E 3 image generation but the last time I was looking it seemed like Stable Diffusion has gotten damn good and is extensible in ways that dall-e is not.

Voice recognition, and TTS output w/emotion OpenAI has the best I’ve ever heard.

Image recognition openAI might lead but the llama4 multimodal stuff is pretty awesome

Anyways I’m just some rando but my observation is that OpenAI better get on that IPO fast unless they have some magic in the pipeline because they are being attacked by competent solutions from every side in a niche that is showing diminshing promise to change everything the father we go.

A lot of their staff will be finished with needing to make any money ever again in their lives if they IPO.

But that will also throw OpenAI into the public sphere of needing to make money quarter after quarter forever, which inevitably leads to them sucking and no longer being innovative. Actually OpenAI is already teetering on this already as others are catching up. Sam and other top guys will leave because they hate people telling them what to do and don’t need it.

Facebook and Deepseek are literally giving it away for free now to kill the competition and in some areas other competitors are producing better products already.

It’s a weird space. I used to think we were on the verge of entering a whole new era of technology but now I think it’s going to be muted. Perhaps AI (as we know it now) eased some tasks and eliminated some BS stuff that used to waste our time, but we’ve not yet completely eliminated most professional jobs - if anything I think we just added more for them to learn and do to remain competitive.

I was in San Francisco last week and got to take Waymo several times. Truly a surreal experience and if properly implemented seems like it’s very much our future.

This. Your data is stored in .md text files so even if Obidian somehow stopped being the best your data is so easy to move around.

Also add to your list mega.nz works for syncing Obsidian across many systems.

Fascinating… I knew some of this and it is indeed troubling.

It seems that Brave’s mission is actually about generating revenue by any method possible (including manipulation of end users) more than anything to do with privacy.

If you’re cool with all that then Brave is for you I guess.

I used mailspring for about 6 months because I love the idea and it looks beautiful. But when you check the forums you see people are complaining about major bugs that seemed to remain unfixed for eternity, developer never comments.

Because spinning disks are a bit cheaper than SSD?

Obsidian has a plug in for this… here is an announcement from the plugin author: https://www.reddit.com/r/ObsidianMD/comments/1bsa6dy/alpha_release_of_my_handwriting_plugin_ink/ (sorry for a reddit link)

Switched from Onenote to obsidian. There was a small learning curve and I had to install some plugins, but I love it. It looks amazing and runs so much faster than OneNote ever did.

Driver support was so dicey. If you had anything even remotely not mainstream, you would be compiling your own video driver, or network driver, or basically left to figure it out for any other peripheral. So many devices like scanners and very early webcams just claimed zero Linux support at all, but you could at times find someone else’s project that might work.

I tried to switch to Linux as a desktop system several times in the late 90s but kept going back to windows because hardware support just wasn’t there yet.