By Thomas | financial enthusiast
My AI diary: September 30 — OpenAI’s Dots and the cheaper Sol model shake up work AI
The launch that feels like a coworker
First thought was, wow, OpenAI finally moved beyond the chat box. I read that they unveiled “dots,” an always‑on agent that keeps working across apps like Slack and Teams without waiting for a human prompt. It’s built on GPT‑6 Astra and can update proposals, do research, analyze data, and even generate documents or code. That shift from reactive assistant to persistent digital coworker felt like a real step toward AI that lives inside our workflows.
Cost vs capability: Sol’s promise
Then came the pricing bombshell for GPT‑6.1 Sol: $2 per million input tokens and $10 per million output tokens, which sources say is roughly one‑fifth of Astra’s standard rate. Despite the lower cost, the model is described as nearly matching Astra on agentic coding, computer operation, and professional tasks. I had to sit with that — cheaper doesn’t have to mean crippled, and it could push enterprises to experiment more freely.
Who’s feeling the ripple
Investors are watching the price/performance battle heat up, while developers see a chance to build apps that rely on background agents rather than constant prompting. Enterprises that need AI to handle recurring tasks — internal research, drafting, support workflows — now have a tangible product to plug in. And workers in admin, sales, ops, and software roles are the most exposed because Dots targets multi‑step digital work, not just simple Q&A.
What the experts are hinting at
Enterprise‑focused coverage noted that ChatGPT Space and Dots could turn assistants into participants in shared team workflows, suggesting AI is becoming a collaborative layer inside companies. Market commentary emphasized that price/performance is now the key battleground, with Sol framed as materially cheaper while retaining much of Astra’s capability. The voluntary U.S. frontier‑model safety accord and the reported delay of GPT‑6.1 Astra for safety reasons are being read as signs that self‑regulation and alignment checks are shaping product roadmaps, not just afterthoughts.
Where this leaves the industry
The bigger picture is that AI is moving from models to systems: the real product is an agent stack combining model, memory, permissions, workplace integrations, and persistent execution. Cost compression is accelerating competition, pushing rivals to differentiate through workflow depth rather than raw benchmark wins. I’m left wondering how quickly the rest of the field will answer with their own persistent agents or cheaper tiers.
What do you think—will a persistent agent like Dots finally make AI feel less like a tool and more like a teammate?
