Simon Smith @_simonsmith

Dots are for doing. Chat is for learning and discussing. Work is what you use if you have no Dot.

@mxstbr another one. In general the chat box at the bottom of the page feels unpolished. For example, on iPhone, once you’ve interacted with it that bar doesn’t go away. It floats above the chat box, disconnected from it. That feels rough. Why not just have the box float as a small circle, open when you interact with it, close completely when you don’t, including the content inside? Like an Intercom chat widget.

keyboard_arrow_left Previous keyboard_arrow_right Next

Wish list: 1. I wish it could load and respond faster. This is using Sites as the backend. It's great, but things seem to load and respond slowly. When I load the page, it takes a few seconds for the content to show. When I click a thumbs-up or thumbs-down icon, it takes a second or two to register. I've tried to have my Dot optimize this but it may be a function of how these are built and integrated. I have the same issue with my Spanish vocabulary app. 2. I wish we could use the Dynamic UI model to generate the content, or an ultrafast image generator model that could add visuals. My Dot adds some, like charts and timelines, but it's pretty limited, and it doesn't seem like the best tool for the job given the existence of the Dynamic UI chat model and image generation. 3. When I click "Discuss," it passes the context to the chat window at the bottom of the screen but gives it the label "Context" (see attached screenshot). This seems like a bug. It should better indicate the text that was passed as context. My Dot says it's doing everything right, so this may be a bug. 4. I'm not sure I'm doing the scheduling right. My Dot is polling every hour for new stuff, but is that the right way to do it? It created a scheduled task for itself. That way, it can poll (I assume it delegates that to a subagent), review, and alert me if needed. But should I be doing that? Or should I put the polling schedule on the Site itself, then have my Dot just check the latest stuff? I'm not sure. I'm not sure if there are guidelines for this or something :) I can ask my Dot. PS: I think it was also you who fixed the alignment issues for custom plugins on the iPhone. Happy to report this looks great on the latest version, so thanks for that!

Updated my ChatGPT news plugin to monitor the news so I don't have to. My Dot now polls for new items from specified sources every hour, adds important stuff to my feed, and alerts me if anything needs immediate attention.

One more minor one. My Dot can't seem to change the icon for the plugin. It can change the sidebar icon no problem. But it can't figure out how to change this, and it seems to be set directly from Sites, because when I try to download this plugin as a zip... the zip is empty. Also, the 1.0.0 version number isn't changing. My Dot says it's setting the version number appropriately, but it doesn't move from 1.0.0, which isn't the plugin's current version.

@mxstbr Thanks. It gave the same advice :)

@a66mike99 It's like Muse, which... oh, right, EU :( Hopefully soon.

@shaunralston One my Dot built me from scratch using ChatGPT Sites as an MCP server + the ChatGPT plugin architecture with extensions. It's entirely custom.

Personal agents taking over my phone.

@ericmitchellai Sorry, for my Dot to pick up, re: phone calls. Confusing Muse/Dots.

I'm thoroughly Dots-pilled. Still a fan of Muse, but probably 75-80% of my AI requests now go through Dots. The primary exception: learning things in chat. For that, GPT-6 Instant with Dynamic UI is unmatched. Just incredible. Some recent Dots work and personal tasks: - Build and update a Spanish vocabulary app - Maintain a document on AI comparisons - Execute tasks I delegate directly from Slack - Monitor threads in Slack and summarize responses - Research a contact and write a brief for a call - Analyze token data and make me a deck to share - Monitor token consumption, alert me if needed - Respond to comments in Pages in Space You might say: "Dude, ChatGPT could do a lot of that already." True. But I no longer have to think about all these threads individually. I just keep firing stuff at my Dots from wherever I am. ChatGPT, Slack, reading a Page in Space. I just throw stuff at it, and it does it, and sometimes it proactively just does stuff based on what it thinks I'd find valuable. I was skeptical when this launched because it's nowhere near as consumer-friendly as Muse. But I think that's part of why OpenAI launched this first to business and pro-tier users. We're willing to suffer a bit for extra capability. I'm starting to see some of the advantages now, and the product also continues to get more polished. Bullish.

Glad to contribute. Biggest frictions: 1. Onboarding isn’t as seamless as Muse. Doubly so if you don’t already have lots of connectors set up. 2. It’s unclear how your Dot relates to the rest of ChatGPT. It breaks your mental model. It’s an assistant inside an assistant. Muse is just Muse. I named my personal one “Chatty” so I just thought of it as the quick conversational entry point to ChatGPT. 3. It’s not clear when it’s doing anything. Muse uses emoji reactions, short text lines (often also with emojis), activity tracking in the sidebar if you want to see it, and the ability to drill-down on each sidebar item to reveal the details. My Dot does stuff and I don’t know what it’s up to, and can’t inspect that. Muse has awesome progressive disclosure. 4. The phone versus normal voice mode paradigm is weird. Especially having to wait for several rings for my Muse to pick up and answer. Very strange. 5. There’s huge value using Dots in Slack. But most consumers won’t be using it there. So there needs to be other channels. Email, iMessage, WhatsApp, etc. Places consumers go more often. 5. Needs to be easier to access my Dot with one click on desktop and web. Clicking “Home” brings me to a new chat screen. Those come to mind right now. Can share more as it comes up.

Yay, Waymo in Toronto! Boo, 2028. @Tesla please come sooner.

Waymo @Waymo

20 hours ago

We’re bringing the magic of Waymo to Toronto! 🇨🇦 In the coming days, you’ll begin seeing Waymo’s all-electric vehicles driving through Toronto’s streets as we lay the groundwork for our ride-hailing service. Our goal is to begin serving riders in 2028. Sign up for future

294 353 4K 562K 364

@pritopian The fact I’m confident this is coming soon because it’s so obviously needed and competitors offer it! Just wish it would hurry up.

Really want an email address for my Dot. Use case: Someone emails me and a bunch of people, I loop in my Dot and ask it to take an action or answer a question within the thread. Just like in Slack. When?

Honestly not sure how we’ll deploy Fable 5.5 in our company if it’s far more expensive than Opus 5.5. We have access to great, lower cost models from Anthropic and OpenAI. This would need to be a leap in some relevant area or we might have to gate access for demonstrated need. Some people always want to use the most powerful model when it’s available, even if it’s more expensive and an equally good model for their use case is available. I think this used to be more justified because bigger, better, newer models tended to deliver demonstrably better output for all requests. But this isn’t the case anymore. Many use cases have been saturated below the frontier.

New Spanish voice is great. I set that as my default so I can switch back and forth between English and Spanish and practice throughout the day. It speaks English with a very slight accent. A lot of the new language-focused voices seem to, which is quite endearing.

Simon Smith @_simonsmith

a day ago

Two welcome improvements to GPT-Live: Less "checking," and a better Spanish voice. Both helpful for my Spanish learning.

0 0 9 2K 0

"Tokens" probably won't remain the primary unit of economics for enterprise AI. OpenAI plans to offer specialized Dots. Think Marketing Dot, Sales Dot, Dev Dot. I can imagine them (and others) selling these for a flat annual fee. Maybe a percent of an equivalent human's salary. Customers get benefits like a digital employee that gets smarter each month as models improve, while OpenAI can grow its margins by improving inference efficiency. All without the headache of model switching, model price changes, and token monitoring. We've gotten so used to tokenomics. But it's broken in a lot of ways. One big one: It's hard for enterprise customers to control and forecast costs, because they can be so variable, depending on pricing that fluctuates constantly, and usage within the organization. Once people get into the habit of delegating to agents, treating them more like colleagues, dropping them into Slack channels, emailing with them, calling them, it seems like one obvious pivot for AI labs to drive increased revenue while giving enterprise customers more predictability is selling agents like employees rather than selling the tokens that power them.

Two welcome improvements to GPT-Live: Less "checking," and a better Spanish voice. Both helpful for my Spanish learning.

OpenAI @OpenAI

a day ago

New voices. Better accents. And less "checking..." GPT-Live-1.1 brings these updates to Voice in ChatGPT, Codex, and dots, with new voices in French, German, Hindi, Mexican Spanish, Korean, and Japanese. Rolling out today to all users. Coming soon to the API.

805 787 11K 1.8M 2K

Becoming a bigger and bigger fan of the new ChatGPT chat model. Just great for explaining complex topics.