Connecting in San Francisco
Hello everyone,
This weekend, I’m in San Francisco and I’m eager to meet new people! Here’s what I’m looking to discuss:
-
Conversations about AI focused on non-technical adoption.
-
Builders involved in exciting projects.
-
Founders creating development tools and infrastructure.
-
Potential limited partners (individuals).
-
A chance to taste the renowned deli board.
Feel free to get in touch with me on Twitter.
Ben’s Bites is made possible by AWS Marketplace
Many ML teams create data pipelines before validating their approach, but the Databricks Data Intelligence Platform allows you to bypass that step with data federation. You can connect directly to Amazon Redshift, experiment on live data, and utilize MLOps tools like CI/CD throughout your machine learning lifecycle. Discover how this works.
OpenAI DevDay was an incredible experience, and their standout release was Dots, OpenAI’s friendly, always-on personal agent.
You can communicate with your Dot via messaging or calls just like you would with a human assistant. You have the option to give it a name and choose an avatar (which is not something you can do with a real person!). Plus, your Dot can proactively send you messages.
At its core, a Dot represents a continuous chat with GPT-6 Astra, serving as an orchestration hub.
Your Dot can take on tasks, initiate multiple chats in Codex/ChatGPT Work to tackle them, and keep you updated on the status. It will remain available in your ChatGPT sidebar, regardless of whether you’re using mobile, web, or desktop. Significantly, your Dot operates independently, meaning you don’t need to keep your computer active at all times.
Pro plan users get access to one Dot, which currently doesn’t consume your usage limits (although the spawned chats for task execution do use your limits).
For those who recall OpenClaw, it relied on multiple Markdown files that we crafted to give our Claws personality and memory. Dots, however, lack a clear mechanism for incorporating such files.
It feels more like a tool-dependent system; for the true magic of Dots, they need to connect with tools that are integral to your life. Start with your calendar and email, but also consider Slack, Notion, Jira, and other messaging platforms. Your Dot is designed to enhance your workflow from there. I suggest checking out this article by Ethan for a better understanding of this transition.
Here are some other highlights from DevDay:
-
GPT-6.1 Sol – This feels like a significant upgrade over GPT-5.6 Sol after GPT-6 Sol was perceived as a step back. Learn more about the model here.
-
Pro 500 – A new plan in ChatGPT, introduced alongside Pro 100 and Pro 200. OpenAI has adjusted the usage limits for each plan to align with a simple multiplier of the $20 Plus plan, translating to multipliers of 5x, 10x, and 25x for $100, $200, and $500 respectively. Users of the $200 plan receive a one-time credit to offset the impact.
-
Sign in with ChatGPT – This feature allows you to use your ChatGPT usage across other AI applications. Developers can apply to integrate this option, enabling users to access AI features through their apps.
-
Plugins – OpenAI’s latest version of “apps inside ChatGPT.” There’s potential for Dots to increase user adoption this time around. Early signs seem promising. For more examples, check out: reminders within ChatGPT and tldraw’s multiplayer canvas.
-
ChatGPT Space – This feature resembles a Notion-style document workspace within ChatGPT, allowing sharing with your agents and team.
-
Ultrafast – This option lets GPT-6 Astra operate at eight times the normal speed, but it uses six times the standard usage. It’s only advisable for those willing to invest significantly in OpenAI’s services.
Google is back on the scene with their latest offering: Gemini 4 Argon, which isn’t available for use yet (true to Google’s usual style). They’ve released benchmark scores indicating it may surpass both GPT-6 Astra and Opus 5.5. Moreover, it’s affordable, with promotional pricing that matches Sol. I’ll wait to discuss this model until we can access it.
Muse for Small Business – Meta has integrated connectors for Shopify, Stripe, QuickBooks, Klaviyo, and more, enabling Muse to assist in running small businesses.
-
Decisions API appears to be OpenAI’s answer to Jev. You can define a question along with possible answers, and GPT-6 Luna classifies the data. This API is approximately ten times faster than typical Luna calls, although it’s not as cost-effective as Jev. OpenRouter has introduced its own version of the Decisions API.
-
Codex cloud now includes reusable environments, which streamline your setup time for every cloud project.
-
Sam Altman shares how he utilizes his Dot to run OpenAI and why he prefers the Ultrafast setting.
-
Cloudflare has launched its new CLI, which encompasses its entire API and is tailored for agents.
-
Opus 5.5 has crafted a custom version of Dots/Grok Bot featuring ChatGPT sign-in, Claude subscriptions, and a cloud computer for every agent.
-
Claude Code is now capable of creating evaluations and optimizing them.
-
2026 in LLMs (so far) is a noteworthy read by Simon Willison.
-
Robinhood plans to enable users to create agents that can execute trades on their behalf.
-
Pi now supports MCP, making a case for this new development after previously stating it would not.
-
Reflect Open is an open-source notes app where every note is a Markdown file accessible by your agents.
-
Enable your agents to visually explain plans, differences, and diagrams, rather than relying on long blocks of text. This skill has recently surpassed 10K stars.
-
Have you abandoned AI experiments? Instead of discarding them, consider extracting what’s valuable from those projects.
-
In one concise paragraph before a meeting, craft a comprehensive report site by the end of it.
-
a16z’s State of Markets II reveals that about 30% of S&P 500 companies report some impact from AI, yet only roughly 2% have established metrics for tracking it.
As we wrap up, it’s an exciting time in the AI landscape. With innovations rolling out continually, there’s so much to look forward to. Let’s connect and dive deeper into these developments together!