AI thread

Advertisements
Been having fun w/ Muse. Not sure it's really doing anything new per se, but it's proactive suggestions are going to expose more and more people to agentic work. For example it's monitoring a few local auctions, ebay and FB marketplace for deals on RAM for my gaming server. The FB marketplace bit is really nice because Facebook's search is just so awful, surfacing so much more than what you are looking for. It also reviews my gmail and aggregates a few things for me to review before sending them to my wife to keep her in the loop (Amazon deliveries and Audible authors she follows on our shared Audible account). I also have it managing a living document as I enter the final year of my EV9 lease, to manage the process of considering keeping it or what alternatives I'd consider moving onto would cost me. I had it login to NVEnergy, request, download and process a year's worth of usage and size a solar / battery setup. Then talked me through my realistic options there.

It also has a news feed that aggregates news somewhat related to both my interests as well as topics I've worked with in the tool.

It's basically what I would have expected Gemini to do months ago.

Much of this stuff others have been able to do for a while, but they have been far less proactive in suggesting things like not just finding a product on Facebook marketplace but offering to contact sellers, haggle a price and arrange a pickup. I haven't let it do that, but crazy to think that it will. A similar task I have Claude do now that it has it's built-in browser is it goes to infinite campus and canvas and aggregates the kids grades, providing me with a clean dashboard w/ links to both of those sites for each classroom so it's easy to talk about schoolwork with them.

I have a referral code that gives you '1 billion muse credits' - whatever that really means. Code: U20F49 Muse — Your Personal AI Agent

If anyone uses it, feel free to post what you end up doing w/ Muse and your code so we can spread the wealth. I think they're going to end up w/ a first mover advantage that's going to print them money thanks to their connected services and hardware.

Pretty close to getting one of their glasses, just feel like the camera setup is still a bit bulky for everyday use. Not sure I want the new audio only version or not.
 
So yeah, I just heard on the news that they're going to intro audio-only glasses
 
Ok tomorrow morning my scanner project will launch to the public at Newington Scanner Summaries

I have the site staged and will update the homepage in the morning to look like this... Now people will know What those Sirens were. :D

Screenshot 2026-09-26 at 6.56.43 PM.webp
 
Lots of AI news today w/ OpenAI's dev day. Dots one of the big visible things.


View: https://youtu.be/uXspbC2srEQ


Dev day site: https://openai.com/index/devday-2026-recap/

A recap from ChatGPT itself:

Here's the "why should a normal ChatGPT user care?" version:

🤖 OpenAI basically announced a new kind of ChatGPT

The biggest announcement is Dots, and the important part isn't the cute avatar. It's that OpenAI is trying to make an AI that keeps working after you stop talking to it.

You can give a Dot a goal, permissions, and access to your apps. It can then work on that goal in the background, using its own computer and browser. OpenAI showed examples involving Slack, email, GitHub, and software projects. OpenAI

That's a pretty substantial change from:
You: "Do this."
ChatGPT: "Here's how."
to:
You: "Take care of this."
AI: goes away and actually does it.

🏢 ChatGPT is also becoming a place where AI and people work together

ChatGPT Space lets a group of people and their AI agents work on the same project. The idea is that the AI isn't just answering questions in a chat window. It can maintain project information, update things, flag problems, and turn work from one person into work for another. Venturebeat

That starts looking less like ChatGPT and more like an AI-powered office.

🧠 OpenAI also made AI substantially cheaper and faster

GPT-6.1 Sol is designed to provide much of the capability of the expensive frontier models at substantially lower cost. OpenAI is also introducing an Ultrafast processing tier that can dramatically speed up AI work. OpenAI Developers

That's less exciting on the surface, but it's important because cheaper/faster intelligence makes all those agents much more practical.

💻 Codex is becoming an employee, not just a coding assistant

OpenAI is pushing Codex toward independently handling chunks of software projects. You can increasingly give it a task and let it investigate the codebase, make changes, run tests, and produce the resulting work rather than sitting there prompting it through every step. Simon Willison's Weblog

💰 And there's a $500/month ChatGPT

This one is less exciting for most people, but it shows where OpenAI thinks the market is going. Pro 500 is aimed at people who use enormous amounts of AI compute, with its highest-speed processing and usage allowances. Business Insider


The really interesting part

Today's announcements aren't really about a better chatbot.

They're about changing the fundamental relationship:

2024:
"Ask AI a question."
2025:
"Give AI a task."
2026:
"Give AI a goal and let it work."
That's why I think the Dots announcement is considerably bigger than it initially sounds. OpenAI is trying to turn ChatGPT from something you open when you need it into something that can remain involved in your work and act on your behalf. OpenAI

And there's a fascinating wrinkle: OpenAI canceled the release of GPT-6.1 Astra's planned launch because of safety concerns just before DevDay, specifically involving the model not reliably staying within its authorized scope. So the company is simultaneously pushing much harder toward autonomous AI while discovering that controlling autonomous AI is one of the hardest problems. washingtonpost.com

That's the real story of today's announcements:
OpenAI isn't merely making ChatGPT smarter. It's trying to make AI independently useful.
 
  • Like
Reactions: b4pjoe
Tip of the spear of how we'll be using computers differently in the future.
I can say without reservation that Microsoft has never been particularly "tip of the spear".

The fact that they've discontinued Copilot PCs confirms that their vision for the future is rudderless.
 
I think this style of PC is developer-focused right now, but our future computers will be focused on local inference and agentic work - just probably different terms and packages of technology once it's more approachable / digestible (probably in a more passive way) for older generations and wider more intuitive usage.
 
What will we become with all of this agentic help?

Less operator, more director. We'll spend less time on the mechanics (chasing down error codes, formatting reports, digging through folders) and more on deciding what we actually want done and judging whether it came out right. Same shift as calculators and search engines: the tedious part gets handed off, and the skill becomes knowing what to ask and spotting when the answer's wrong.
 
Same shift as calculators and search engines: the tedious part gets handed off, and the skill becomes knowing what to ask and spotting when the answer's wrong.
This begins to sound a lot like trial and error rather than thoughtful design -- very much like Adobe and Microsoft. Also known as the "Monte Carlo Method".

It also becomes a liability, as AI (especially as commonly applied to software development) is less about creating innovative approaches than it is about regurgitating solutions it has already seen.
 
Agreed, it mostly regurgitates known solutions. That's the point. Nobody needs an innovative fix for a driver error or a creative way to format a report. Let it handle the solved problems so people can spend their time on the ones that aren't.
 
That's the point at which innovation ends, and there's no longer any point to software engineering other than swatting bugs that result from misapplication.

Addressed by what you didn't quote. This cycle is as old as computing: compilers, libraries, frameworks each automated the tedious part, and the human work moved up a level. Innovation didn't end; it shifted. And people still make new compilers, libraries or frameworks.

Nobody's forced to hand anything off. Anyone who wants to keep doing it by hand still can, and people who couldn't do it at all now can.
 
Whether from others or our own mistakes, we've all learned that lesson one way or another.
I'm not sure that the school of hard knocks is the quickest or surest education to find your way.

There are many examples in the headlines of people getting burned multiple times and not wavering in their pursuit of stupidity.
 
I'm not sure that the school of hard knocks is the quickest or surest education to find your way.

Good thing not everyone will have to learn the hard way then. The same tools can explain why something broke.

There are many examples in the headlines of people getting burned multiple times and not wavering in their pursuit of stupidity.

This is true of literally anything, though.

Happy to agree to disagree, but I'm already enjoying the dividends personally and helping others at work and at home do the same.
 
  • Like
Reactions: telstar_1

From a time when people could still laugh

Gas Price Info With National Gas Temperature Map