Today we’re introducing Gemini 4 Argon.
It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.
・Carry my corrections forward across chats and days, so I don’t have to repeat the same feedback.
・Continue with work that can proceed independently, and ask necessary questions together rather than stopping for each one.
・Make it clear to both users and dots which files are current and which decisions are settled in a Space.
My main takeaways from DevDay:
1. dots
This might have been the announcement I was most interested in.
The shift from asking AI to do something every time to giving it ongoing work feels more important than simply having more Agents.
What I really want to see is how much less time I have to spend managing AI itself. Being able to use more AI matters less to me than having to manage it less.
2. GPT-6.1 Sol
I'm more interested in the economics of long-running work than the benchmark improvements themselves.
If AI is going to keep working with a long context, cheaper cached tokens — reusing input it has already processed — can make a big difference.
But API pricing isn't the only cost.
・If I have to repeat instructions,
・review the work every time and fix the output,
・then my time is part of the cost too.
So even if a model costs more per token, it can still be cheaper if it gets the job done in one go.
3. Agents API
I see this less as an "API for building Agents" and more as infrastructure for handing off long-running work.
As AI starts working for hours rather than minutes, what happened, where it stopped, and how it recovers become more important.
The metric I want to look at is how long I can leave it alone and still trust the work.
4. Plugin Extensions
This feels very practical. Just because more of my work involves AI doesn't mean I want more SaaS tabs.
If I can use services directly inside GPT, GPT starts becoming less of a place to ask questions and more of a place where work begins.
Fewer screens, less switching, less unnecessary overhead.
5. Sign in with ChatGPT
This could be a pretty significant change for people building AI products.
I've never thought it was a particularly good structure for every SaaS company to pay for model usage, meter it, and then resell that usage to customers.
In addition to identity through Sign in with ChatGPT, some participating services allow Plus / Pro users to use their own ChatGPT plan allowance. That lets product companies focus more on the value they actually create.
6. Plugins / OpenAI Marketplace
This gets interesting when combined with Sign in with ChatGPT.
Plugins can be discovered and used through a shared Directory. OpenAI Marketplace, on the other hand, is a separate mechanism that allows eligible Enterprise customers to apply part of their existing OpenAI commitment toward approved partner products.
OpenAI is moving beyond models into identity, AI execution infrastructure, distribution, and enterprise purchasing channels. Smaller companies may no longer need to build all of these from scratch.
Simply "using AI" becomes even less of a differentiator.
In the end, what matters is what problem you actually solve.
7. Usage & Pricing
The direction of pricing.
I'm not particularly interested in paying less for AI. What I care about is how much more valuable one hour of my own time becomes because of what I spend on it.
I'm starting to think of AI spend less as a software subscription and more as an investment in productivity.
-
For me, and probably for society too.
I want to value everything we've learned and built up to today, and focus on what we can do with it now and from here.
❀
The number of times I have to step in with corrections, additional instructions, or decisions.
If that number goes down day by day, dot is growing. If it doesn’t, it’s just “remembering”—not yet ready to be trusted with work.
I’ll start by tracking this for three days, starting today.
◢◤OpenAI「Sign in with ChatGPT」— ChatGPTの利用枠を、外部ツールへ。
OpenAIが「Sign in with ChatGPT」を発表しました。対応する外部サービスへChatGPTアカウントでログインし、PlusやProなどで持っている利用枠を、そのサービス内でも使える仕組みです。
一般的な「Googleでログイン」などと違うのは、本人確認だけではなく、AIを使うための利用枠まで一緒に持ち込めること。
外部サービス側から見ると、利用者が試す段階からAI利用料をすべてサービス側で負担する設計だけではなく、利用者自身のChatGPT契約を使ってもらう選択肢ができます。
AI製品環境の競争では、モデルの性能や価格に加えて「最初に使い始めるまでの費用と手間を誰が持つか」も、製品設計の一部になっていくことを感じる発表です。
【憶えておきたい用語・英語表現】
Bring your own subscription:利用者が既に持つ契約・利用枠を外部サービスでも利用する考え方。
今回の「Sign in with ChatGPT」は、認証だけではなく利用枠の持ち込みまで含む点が重要。
You can now use your ChatGPT subscription directly in over 16 partners products. No little rules, you can just use all your included usage right there. This includes things such as Devin, OpenCode, Notion, and many more.
Sign in with ChatGPT and go.
Still, paying for AI is also an investment in how much value you can get out of one FTE, so I think this is a little harsh for people who are already overwhelmed with day-to-day work and rely on AI just to keep going.
Because in most cases, people who just use Astra 6 Max for everything still have to keep working anyway.
This announcement shows how agent permissions are expanding beyond application settings into the execution infrastructure.
What stands out is that the architecture goes beyond asking agents to follow rules: it enforces those rules at a layer the agents themselves cannot modify.
We need to consider not only how to define what we delegate to agents, but also how to enforce those decisions as actual execution boundaries.
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. https://t.co/ugYWQ1MyRi