Austin

@tszzl Or a poor state

the value of this technology will mostly not be captured by its inventors, the labs, or even the chipmakers, but rather will be captured by the consumers as surplus. these are highly competitive markets without any natural monopolistic effects like many other technologies before it, machine intelligence democratizes abilities previously only available to the wealthy, in this case by commoditizing the services of the white collar elite who mostly live in rich countries it’s not that there are no programmers, it’s that really anybody can make software now now so the “rents” of the “human capital” of knowing how to write JavaScript for example should shrink dramatically this will reduce the inequality between countries: services that previously required lots of human capital now require chatbot subscriptions at worst, or may even be given away for free you can receive medical advice worthy of a $1000/hr American specialist doctor likely for free while living under a thatched roof in eg Papua New Guinea somewhere while I think Americans have plenty of reason to be excited by AI, I would be more excited as someone in a poor country

@kurtbuhler Just to calibrate, what's an example where subagent use is annoying?

Opus 4.6 in Claude Code is using subagents a little too frequently, it's getting a little annoying for certain work and required quite a few modifications of the user CLAUDE.md to fix.

Been waiting for Wispr Flow on Android? Win stuff on this 🎰 waitlist: wisprflow.ai/waitlist?AUSTI…

@chatgpt21 They lost their Claude code subscriptions

Where’s Xais genie? Elon said 2 months 6 months ago…

@MachinesBeFree Whatever's in those Epstein files must be really bad

@yooo_chip @mckaywrigley two of my favorite drugs for sure

@mckaywrigley @austinsense as a person who loves to talk about claude and GLP-1s, do you care to weigh in here?

claude is ozempic for knowledge work

Underappreciated unhobbling

@koltregaskes @claudeai Also playwright MCP 🔥

@claudeai Can Code control thw browser though, like your Chrome plugin?

Claude in Chrome is now available to all paid plans. We’ve also shipped an integration with Claude Code.

@koltregaskes Agentmail | The Email Provider for AI Agents share.google/Kg2ju6DmTKUH1f…

Update: I have Claude Code, Gemini CLI and Codex CLI also set up in the Obsidian workspace and we're all planning how to work together and set each agent up, e.g. Nano Banana for Gemini, deep research for Code and manager for Claude. I wish I could get them to talk directly to each other though. I'm currently getting them to write in .md files in respective ai assistant folders.

this acfs is really something @doodlestein

@DaveShapi @joshwoodward Yes, this

What I really want is a Digital Employee. Something that learns the job over time by developing more context and updating a version controlled library of tasks and tribal knowledge.. Probably be run by several agents. Context management agent. Version control task agent. User interface agent.

Imagine it's early 2026, and you're using a Gemini MacOS desktop app. What features *must* be there for you to love it? Anything you really want to see in the desktop app specifically?

@kurtbuhler @JohannesVink Just finished a Claude code training today, asked if we could get a testimonial and that was verbatim what one of the attendees said. 😂

I do agree. I'm very skeptical and very critical about AI... but I agree. Esp. on the non-engineer PoV. The change that's coming is just... unfathomable. I literally cannot stop thinking about it. Every hour of every day. Over the last few weeks I introduced Claude Code to some friends in different domains -- academic research, pharma, medicine.. the reaction after 2-3 days is always about the same: "Holy fuck."

The abstraction and convenience that the self service platforms used to provide is feeling more and more like a cost and constraint when using agents for development.

@kurtbuhler @JohannesVink It's hard to convey to people that if you're in these new AI tools everyday, as a non-engineer, you start to develop some interesting, valuable intuition about where the future is headed. Extrapolate out a few years, and the world looks very different.

Example #2: My site data-goblins.com is hosted via squarespace because I'm not a web developer. I used a CMS because of the convenience, accepting more limitations and restrictions, like less customization and no source control (eek!). However, again, I asked myself -- why should I tolerate these limitations? Can't agents tools provide the same convenience that the platform does? So I am switching, migrating my entire site to a different open source platform that gives me full code-first control. I fleshed out a few agent tools to cater to the things I do and want, tailored my design system to better agent use, and it's done; I'm now extending it with things that weren't possible before. Again, the self-service platform of squarespace used to provide me the convenience of site hosting, design, and content delivery. However, this layer adds complexity for an agent to do what I want; adjusting some code, updating a diagram etc. (I don't use AI for content creation at all, to be clear). So I decided to cut that complexity; simplify to code and markdown so I have more flexibility, more automation, and source control. @leerob wrote about this recently; I relate to it very much: leerob.com/agents

@DaveShapi a bittersweet blend of accomplishment and sadness

don't ask me about what true pain feels like

@danshipper Sorry I was asking you to be the keynote at our 500 person 2026 kickoff event for Birmingham AI

something must be done about this absolute abomination

@yooo_chip Sir, this is very forward of you

may we meet?

@kurtbuhler This is the way

I've been using the Fabric CLI a lot. Reflecting on it, I honestly think this is the most impactful feature for me from Fabric OR Power BI in recent years. I really do mean that. If you want an introduction to the Fabric CLI, I've written about it here: tabulareditor.com/blog/agentic-o… Since I discovered the Fabric CLI, I find that I'm using it daily. TBH, I don't think I've EVER used ANY Power BI feature daily. To be clear, I don't execute commands most of the time; I give it to coding agents (I favor Claude Code) who do it for me. Incrementally over the months I've built up context files and examples, and compiled those in a few skills. Here's a few examples of things I've done with it, just in my own tenant: 1. I got a notification that a model refresh failed. I was at a conference and I needed that model for a demo, but I didn't have time to look into it. I just sent Claude Code after it using the Fabric CLI, asking to inspect the refresh history, model, notebooks, and data to figure out the issue. In a few minutes it crawled through the workspace and found the issue in one of the notebooks, fixed it, and then re-ran each notebook in sequence and refreshed the model, confirming if the refresh succeeded and querying it to validate. 2. I saw my Fabric Trial ticking down to 3 days and realized I needed to migrate all my trial workspaces. I can't automate this deterministically because each workspace needs a separate human decision: migrate to PPU, migrate to F SKU, migrate to Pro, or archive and delete. So I created a re-usable prompt and with a few parallel agents went through and did it in 20 minutes. I wrote about that here: lnkd.in/eufDkMkh -- my trial did get renewed though :P 3. I use it to facilitate CI/CD during agentic report development. This is something I've shared briefly ( lnkd.in/eqBTRDCx ) and started documenting in a longer video and article, but illness and injury in the last month has kept me from finishing it... IYKYK. Of course the Fabric CLI is also amazing for deterministic automation scenarios too, which many have blogged about like Peer Grønnerup ( lnkd.in/e-vJK52X ) and Kevin C. ( lnkd.in/ek4YUYiR ) If you were to tell me last year that 2025 was the year I'd fall in love with the command-line, I'd have laughed at you. "I'm a designer! I practically live in Figma and Excalidraw." And yet, here we are -- Claude Code and the Fabric CLI changed everything, for me. That's pretty cool.

@koltregaskes @GeminiApp It's out but it took a lot of clicking on different menus to figure out how to turn it on

Ow, I just had a @GeminiApp-like answer on my Google Nest Max speaker. Different voice, different text formatting on the display, more intelligent answer. I can't get it to do it again, though. Gemini at Home must be imminent then. Very excited. Google Assistant is still one of my most used AIs because it's on my speakers.

@maxmarchione *sometimes

Personal health snippet: - 6pm today feeling dead tired, headache, brain dead - 7pm did 20min of yoga nidra - 8pm played 45min of soccer with superpower team - 9pm feel completely alive, better than ever sometimes the answer to fatigue is to do more

@shreyas I feel seen

Being highly competent but politically unsavvy relative to your peers — whether you’re at a fast-growing startup or a big company — might prevent you from reaching your full potential in conventional terms (e.g., org size, title, etc.), yet it keeps you closely connected to who you truly are. Over time, this builds a quiet, intrinsic confidence that many of your peers will lack despite their big titles and large orgs. What you do with that confidence is up to you, but in many cases it allows you to make more authentic career choices your former peers say they would love to make “if only the timing were right.” Of course, what appears to be a timing problem is really a confidence problem — the quiet, authentic kind of confidence. ~ Now, this isn’t saying that being politically savvy is “bad” or that being politically unsavvy is “good.” A fundamental problem in our society these days — and in most online content — is the pressure to believe all humans must be the same: same thoughts, same personalities, same superpowers, same life goals. In reality, people are different, will always be different, and there is no single “ideal formula” everyone must follow. Every parent with more than one child and everyone with siblings knows this truth: people are different. No amount of forcing, lecturing, modeling, or incentivizing will change that. Yet our societal conditioning runs so deep that we come to believe everyone else should not only resemble us but should also adopt our beliefs, our priorities, and our way of thinking — as if our own ideas were the universal template everyone else must conform to.