OpenAI copia Grok con i dots? GPT-6.1 sfida Opus 5.5
A comprehensive analysis of OpenAI's DevDay announcements, including new AI agents called 'Dots,' the GPT-6.1 model, and pricing changes, alongside broader trends in AI development from competitors like Anthropic and open-source projects. The speaker critiques the announcements as largely incremental improvements rather than groundbreaking innovations.
Summary
The transcript covers OpenAI's DevDay presentation, which introduced several new features and models. 'Dots' are cloud-based AI agents that can work independently and communicate with each other, positioned as OpenAI's answer to GrokBot. These agents can be controlled via voice commands, phone apps, or desktop interfaces and have access to browser automation and computer control capabilities. The speaker notes these are not particularly novel, as similar agents like OpenInterpreter, Hermes Agent, and GrokBot already exist.
A notable feature is 'Spaces,' collaborative workspaces where multiple users and AI agents can work together simultaneously on documents, infographics, and images. The speaker highlights a significant developer advantage: the ability for users to log in with their ChatGPT subscription rather than requiring developers to manage API costs, which benefits low-budget projects and open-source software.
On the model front, OpenAI released GPT-6.1 SOL as its flagship model, offering intelligence comparable to Astra at one-fifth the price. This sparked a pricing war with Anthropic, which released Opus 5 and Sonnet 5.5 with similar competitive advantages. The speaker notes confusing pricing tiers with multiple reasoning modes (High, Max, Ultra Fast), criticizing the complexity and questioning whether the cost-benefit tradeoffs are justified.
The speaker discusses the new era of 'System One models' (decision models) for real-time automation like browser control, acknowledging OpenAI's entry into this space alongside competitors like DeepSeek's Jeev and Rizzo Flow. Additionally, OpenAI is expanding its marketplace with plugins that integrate third-party apps directly into ChatGPT, enabling new business models where applications live within ChatGPT rather than as standalone apps.
The analysis includes observations on open-source developments, particularly Xiaomi's MiLM 2.6 Pro (a multimodal model handling text, image, video, and audio) and Cen Image 2.1 (an image generation model runnable locally). The speaker predicts Chinese models will soon catch up to frontier models through knowledge distillation and reinforcement learning techniques.
The speaker concludes that while pricing continues to drop, token consumption will increase due to multi-agent systems communicating and generating content simultaneously, resulting in higher overall costs despite per-token reductions. The overall assessment is that DevDay lacked groundbreaking innovation, instead focusing on productizing existing concepts and refining current offerings.
Key Insights
- OpenAI's Dots are not genuinely novel—they are a cloud-based reproposition of agents that already exist like OpenInterpreter, Hermes Agent, and GrokBot, with the main difference being they run on OpenAI's cloud rather than locally
- The new subscription model allowing users to authenticate with ChatGPT directly within third-party applications, so users spend their own subscription credits rather than developers paying API costs, is the most valuable feature for developers with limited budgets
- Despite token costs dropping significantly (GPT-6.1 SOL costs $2 per million input and $10 per million output tokens, five times cheaper than Astra), overall AI spending will increase because multi-agent systems communicating simultaneously will massively increase token consumption across the board
- OpenAI's pricing tier system with multiple reasoning modes (High, Max, Ultra Fast) is overcomplicated and counterintuitive—the speaker observed cases where 'High' effort settings produced better benchmark scores than 'Max' effort settings while costing less
- OpenAI has entered the System One decision model space, similar to competitors like DeepSeek's Jeev and Rizzo Flow, enabling real-time decision-making for browser automation and robotics, marking a new era of models that make decisions rather than just generate tokens
Topics
Transcript
[0:01] It's honestly hard to keep up these days because they're releasing a mind-blowing amount of new features. In the span of 10 days, they’ve released something like four new frontier models . Yesterday was OpenAI's DevDay, where they presented so many new things. Among them are "Dots," which are nothing more than a blatant copy of GrokBot; they are AI agents and assistants that live in the cloud, talk to each other, have access to their own computer, and run tasks in the [0:33] background—something Meta is also doing with its Muse agent. Fun fact: as soon as Elon Musk saw the announcement of OpenAI's Dots, he bought the domain dots.com, paid 15-20 million for it, and now if…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Simone Rizzo
Qwen 3.8 Max, DeepSeek e Kimi: cosa sta succedendo davvero nell'AI
Chinese AI laboratories are releasing frontier open-source models weekly at 1/3 to 1/4 the cost of American models, forcing dramatic token price depreciation and shifting the AI industry's competitive focus from model intelligence (now a commodity) to hardware infrastructure, chips, and software optimization. This geopolitical competition is reshaping business models across the industry, with major companies pivoting toward open-source releases and custom hardware development.
gpt 5 6 sol, grock 4 5 e muse spark 1 1
A comprehensive review of recent AI model releases including Meta's Muse Spark 1.1, Elon Musk's Grock 4.5, and OpenAI's GPT 5.6 family (Sun, Earth, Moon variants). The speaker analyzes performance benchmarks, pricing, and capabilities, concluding that while new models are competitive, none have achieved a significant leap beyond current frontier models like Claude Fable 5.
Tutti parlano di Loop Engineering... ma nessuno te lo spiega così
The video traces the evolution of AI interaction paradigms from Prompt Engineering through Context Engineering, Harness Engineering, to the newest Loop Engineering approach. Loop Engineering involves wrapping autonomous AI workflows in iterative loops that self-improve toward defined goals without requiring manual intervention between steps.
DeepSeek ha appena reso TUTTI gli LLM più veloci
DeepSeek's new Spark technique uses semi-autoregressive speculative decoding to accelerate LLM inference by 51-400% without quality loss or model retraining. By combining a fast parallel draft model with an efficient verification process, Spark achieves higher token acceptance rates than competing methods like Eagle 3 and Flash, enabling faster inference on consumer hardware.
GLM 5.2 gira in locale quantizzandolo ad 1bit! #intelligenzaartificiale #aiagent
Researchers successfully ran the 744-billion parameter GLM 5.2 model locally on a Mac Studio M3 Ultra using dynamic quantization, compressing it from 810 GB to 223 GB. The 1-bit quantized version maintains 76.2% accuracy while being 86% smaller, and performs comparably to closed-source models like Claude Opus and GPT-5.5.