🤝In Partnership: Where fintech leaders get the full picture

🤿Deep Dive: OpenAI releases GPT-5.4 unifying reasoning, coding, and native compute use

🤝Powered by: AI browsing, built for trust

⚒️ Tool Spotlight: Build app with multi-model mode

🤿Deep Dive: Dario Amodei breaks silence on Pentagon blacklisting and leaked memo

⚒Tool Snapshots: Tools for AI, no-code, and productivity

🖼AI Art: Examples of great and trending AI art

GPT-5.4 introduces tool search in the API, replacing the practice of loading all tool definitions upfront by letting agents look up only the tools they need at the moment of use reducing total token usage by 47% across 36 MCP servers while maintaining the same accuracy, making large-scale agentic workflows faster and significantly cheaper.

The model's native computer-use capabilities allow agents to write code via libraries like Playwright and issue direct mouse and keyboard commands in response to screenshots, with developer-configurable confirmation policies for adjusting safety behavior to match different risk tolerances across use cases.

GPT-5.4 Thinking in ChatGPT now outlines an upfront plan before responding to complex queries, letting users redirect or adjust mid-response without starting over, while also thinking longer on difficult tasks without losing coherence or context from earlier in the conversation.

On factual accuracy, GPT-5.4's individual claims are 33% less likely to be false and full responses are 18% less likely to contain any errors compared to GPT-5.2, with /fast mode in Codex delivering up to 1.5x faster token velocity for developers who need speed without sacrificing model intelligence.

On GDPval, it matches or exceeds industry professionals in 83% of comparisons across 44 occupations, scores 87.3% on investment banking spreadsheet tasks (vs. 68.4% for GPT-5.2), and surpasses human performance on OSWorld-Verified at 75% versus the human baseline of 72.4%. Available now to ChatGPT Plus, Team, and Pro users as GPT-5.4 Thinking, and via the API as gpt-5.4, with GPT-5.4 Pro rolling out to Pro and Enterprise plans.

Anthropic , which signed a $200 million DoD contract in July as the first AI lab to run models on classified networks has signaled it will challenge the designation in court even as it works to de-escalate, with tensions rooted in the company's refusal to allow Claude to be used in fully autonomous weapons or domestic mass surveillance.

Amodei's leaked memo, sent to staff just hours before the official supply chain risk notice arrived, was later described by Amodei himself as written after a difficult day for the company and not reflective of a considered or refined view of the situation, a characterization he repeated to The Economist while stopping short of a direct apology to President Trump.

Anthropic's core tension with the Pentagon stems from the company's AI use restrictions around fully autonomous weapons and domestic surveillance, drawing a line that put its $200 million classified-network contract at risk and exposed the wider challenge AI safety-focused labs face when navigating defense relationships.

Amodei said Anthropic is doing everything it can to de-escalate while simultaneously pursuing legal action to block the designation, framing the court challenge as necessary to prevent a chilling effect on the company's broader business and government partnerships.

🤖 Aident AI Beta 2 - AI-powered automation for efficient workflow management.

📸 Heywa - Transforming prompts into dynamic, interactive visual stories.

🔒 Golf - Centralized AI and MCP server security and compliance control.

📚 Parsewise - AI-powered document analysis and cross-referencing without coding.