2026-02-14
We've seen how much speed affects people's model preferences recently (e.g. the arena @swyx is running), so I think Codex Spark will be well-received. Also interesting that this initial release is a step towards combining long-horizon and real-time agents, including delegating to [image]
ZDNET
OpenAI debuts a research preview of GPT-5.3-Codex-Spark, a smaller version of GPT-5.3-Codex that it claims generates code 15 times faster, for ChatGPT Pro users
2026-02-13
We've seen how much speed affects people's model preferences recently (e.g. the arena @swyx is running), so I think Codex Spark will be well-received. Also interesting that this initial release is a step towards combining long-horizon and real-time agents, including delegating to [image]
ZDNET
OpenAI debuts a research preview of GPT-5.3-Codex-Spark, a smaller version of GPT-5.3-Codex that it claims generates code 15 times faster, for ChatGPT Pro users
ZDNET's key takeaways — OpenAI targets “conversational” coding, not slow batch-style agents. — Big latency wins: 80% faster roundtrip, 50% faster time-to-first-token.
2026-02-12
Microsoft is ubiquitous in businesses, yet Suleyman believes in 1-1.5 years AI will automate most of its customers' knowledge work. What's the pivot for Microsoft if companies don't need as many seat licenses, because they have fewer people, and AI agents don't need office apps,
Financial Times
Mustafa Suleyman says Microsoft is pursuing “true self-sufficiency” in AI by building models for enterprise and health care and reducing its reliance on OpenAI
The GLM-5 benchmark chart doesn't compare the model to Opus 4.6 or GPT-5.3, but is still impressive and, on the heels of Kimi K2.5, suggests China is very close to the frontier in many (but not all) domains. And with video, looking at Seedance 2, China might be ahead.
Reuters
Z.ai says it will raise prices by at least 30% for new GLM coding plan subscribers to accommodate surging demand for its AI coding tools
The GLM-5 benchmark chart doesn't compare the model to Opus 4.6 or GPT-5.3, but is still impressive and, on the heels of Kimi K2.5, suggests China is very close to the frontier in many (but not all) domains. And with video, looking at Seedance 2, China might be ahead.
Z.ai
Z.ai launches GLM-5, saying its flagship open-weight model has “best-in-class performance among all open-source models” in reasoning, coding, and agentic tasks
We are launching GLM-5, targeting complex systems engineering and long-horizon agentic tasks. Scaling is still one of the most important ways …
We've seen how much speed affects people's model preferences recently (e.g. the arena @swyx is running), so I think Codex Spark will be well-received. Also interesting that this initial release is a step towards combining long-horizon and real-time agents, including delegating to [image]
ZDNET
OpenAI launches a research preview of GPT-5.3-Codex-Spark, a smaller version of GPT-5.3-Codex that it claims generates code 15 times faster, for Pro users
ZDNET's key takeaways — OpenAI targets “conversational” coding, not slow batch-style agents. — Big latency wins: 80% faster roundtrip, 50% faster time-to-first-token.
2026-02-11
Substantial deep research update in ChatGPT: - Can specify apps and sites - Makes a plan, which you can modify - Puts output in a nice report viewer The report viewer looks slick, similar to the spreadsheet viewer and PowerPoint viewer OpenAI previously released, and leading me
The Decoder
OpenAI updates ChatGPT's deep research tool with GPT-5.2, a full-screen report view, an option to focus research on specific websites, and search interruption
The feature now runs on the new GPT-5.2 model, as OpenAI announced on X. A key addition is that users can connect apps to ChatGPT and—potentially very useful—search specific websit...
Substantial deep research update in ChatGPT: - Can specify apps and sites - Makes a plan, which you can modify - Puts output in a nice report viewer The report viewer looks slick, similar to the spreadsheet viewer and PowerPoint viewer OpenAI previously released, and leading me
Matt Shumer
GPT-5.3-Codex and Claude Opus 4.6 can meaningfully contribute to the improvement of AI models, a sign of what's coming for most knowledge work within five years
Think back to February 2020. — If you were paying close attention, you might have noticed a few people talking about a virus spreading overseas.
Substantial deep research update in ChatGPT: - Can specify apps and sites - Makes a plan, which you can modify - Puts output in a nice report viewer The report viewer looks slick, similar to the spreadsheet viewer and PowerPoint viewer OpenAI previously released, and leading me
Sources
Q&A with Fidji Simo on ChatGPT ads, OpenAI's efforts to ship a new model soon to end Sam Altman's Code Red, Anthropic's Super Bowl ads, Sora, Codex, and more
How ads in ChatGPT will work, what will end the Code Red, those Anthropic attack ads, working with Sam Altman, and much more...
2026-02-06
OpenAI isn't saying it, but Frontier to me is basically like a platform for digital employees. Or, like OpenClaw for organizations, where you create autonomous digital agents, give them skills and tools, and oversee their work. We're truly going through an evolution into an era [image]
The Verge
OpenAI launches Frontier, an AI agent management platform that provides shared context, onboarding, and permission boundaries, for “a limited set of customers”
2026-02-05
Again, I don't want ads in my chatbot either, but buying Super Bowl ads to criticize ads in chatbots feels... off. It's an implicit moral statement about the value of time spent in chatbots versus time spent watching sports.
@sama
Sam Altman says Anthropic's Super Bowl ads are funny but “dishonest”, and Anthropic serves a “product to rich people” while OpenAI is “committed to free access”
First, the good part of the Anthropic ads: they are funny, and I laughed. But I wonder why Anthropic would go for something so clearly dishonest. Our most important principle for a...
Again, I don't want ads in my chatbot either, but buying Super Bowl ads to criticize ads in chatbots feels... off. It's an implicit moral statement about the value of time spent in chatbots versus time spent watching sports.
Wall Street Journal
Anthropic plans a 30-second ad during the Super Bowl that parodies the prospect of intrusive ads in AI conversations, and will also air a 60-second pregame ad
Ads are coming to AI.Taylor Herzlich /New York Post:Super Bowl commercial sees Anthropic mock OpenAI for bringing ads to ChatGPTJames Peckham /PCMag:Anthropic Says No Ads on Claude...
2026-01-31
So, so close to massive automation. Put these specialists into an agent swarm coordinated by an orchestrator, and...
TechCrunch
Anthropic expands its agentic plugins, which let enterprise users automate department-specific workflows, from Claude Code to its new general-use tool Cowork
Earlier this month, Anthropic launched Cowork, a new agentic tool designed to take the benefits of its AI coding assistant Claude Code …
2026-01-30
So, so close to massive automation. Put these specialists into an agent swarm coordinated by an orchestrator, and...
TechCrunch
Anthropic expands its agentic plugins, which let enterprise users automate department-specific workflows, from Claude Code to its new general-use tool Cowork
Earlier this month, Anthropic launched Cowork, a new agentic tool designed to take the benefits of its AI coding assistant Claude Code …
2026-01-28
Very cool new tool from OpenAI for writing scientific papers. Two thoughts came to mind. First, while not a new model, this can help accelerate science. Second, this could be a precedent for other OpenAI document editing tools, such as for docs and sheets and presentations.
MIT Technology Review
OpenAI launches Prism, a free cloud-based LaTeX editor that embeds GPT-5.2 to assist in scientific paper drafting and citation management
OpenAI just revealed what its new in-house team, OpenAI for Science, has been up to. The firm has released a free LLM-powered tool for scientists called Prism …
2026-01-27
I tried Kimi K2.5 in Agent Swarm mode today and can say that the benchmarks don't lie. This is a great model and I don't understand how they've made something as powerful and user-friendly as Agent Swarm ahead of the big US labs.
Kimi
Moonshot says Kimi K2.5 builds on K2 with “pretraining over ~15T mixed visual and text tokens” and “can self-direct an agent swarm with up to 100 sub-agents”
Today, we are introducing Kimi K2.5, the most powerful open-source model to date.
New Qwen model significantly improves at Humanity's Last Exam with tool use and now appears to be SOTA here by a large margin? Main innovation driving this seems to be a non-parallel approach to test time scaling where the model avoids redundant thinking. On the surface the [image]
Qwen
Qwen releases Qwen3-Max-Thinking, its flagship reasoning model that it says demonstrates performance comparable to models such as GPT-5.2 Thinking and Opus 4.5
· QwenTeam丨Translations:.体中文 — Introduction# — We present Qwen3-Max-Thinking, our latest flagship reasoning model.
Very cool new tool from OpenAI for writing scientific papers. Two thoughts came to mind. First, while not a new model, this can help accelerate science. Second, this could be a precedent for other OpenAI document editing tools, such as for docs and sheets and presentations.
MIT Technology Review
OpenAI launches Prism, a free cloud-based LaTeX editor that embeds GPT-5.2 to assist in scientific paper drafting and citation management
Accelerating science writing and collaboration with AI.
2026-01-26
New Qwen model significantly improves at Humanity's Last Exam with tool use and now appears to be SOTA here by a large margin? Main innovation driving this seems to be a non-parallel approach to test time scaling where the model avoids redundant thinking. On the surface the [image]
Qwen
Qwen releases Qwen3-Max-Thinking, its flagship reasoning model that it says demonstrates performance comparable to models such as GPT-5.2 Thinking and Opus 4.5
2026-01-22
Charging per ad view is interesting for a few reasons, one of which being that Altman and Ive said they're trying to get people off screens with OpenAI's hardware. So not sure a CPM model will work with those future form factors.
The Information