AI Products news and analysis on Web Pulse: 44461 curated articles tracking the latest AI Products developments, tools, and research, updated continuously from vetted sources.
DeepSeek AI's model card for DeepSeek V4.1 Flash reports a global KV cache footprint of roughly 890 bytes per token, about a quarter of the 4x-larger footprint of predecessor DeepSeek V4 Flash and a 437x reduction versus…
OpenAI released GPT Image 2.5, an updated image generation model available in two variants — Flare, a faster, lighter version, and Sunburst, a heavier model for demanding generations — accessible through ChatGPT, Codex, …
Next AGI released Nex-N2.5 Mini, a multimodal agentic model built on a Qwen3.5 mixture-of-experts backbone, under an Apache 2.0 license on Hugging Face, requiring two 80GB-class GPUs to run. In hands-on testing deployed …
Nex AGI's Nex-N2.5 Mini, the smaller model in the company's N2.5 agentic family, requires two 80GB-class GPUs such as H100s and consumed roughly 66GB of VRAM per card in testing on RunPod, according to a deployment guide…
Two Foundry released TrueForge, an MIT-licensed, model-agnostic open-source agent harness that adds sandboxing, code mode, human-in-the-loop approval gates, and context management to production AI agents. TrueForge runs …
Vercel is offering Tako Search free exclusively on its AI Gateway through September 30th. Tako Search is a model-agnostic search tool that works with any model on AI Gateway, combining a curated knowledge graph with the …
Together AI launched public preview of preemptible compute for Together GPU Clusters, offering interruptible GPU nodes at a flat 50% of the on-demand rate on Kubernetes clusters in all regions. Preemptible nodes use the …
IMG.LY released CE.SDK v1.82, adding a full captions workflow to the Android video editor and clip-to-clip video transitions to iOS. The web editor gains customizable timeline controls, new Tabs and Stepper input UI comp…
Vercel now supports building and deploying long-running, tool-using agents with the OpenAI Agents API, with OpenAI managing the agent loop and session state while Vercel hosts the application and connects each session to…
A reference published by Digital Applied proposes a field-level policy for AI tool responses, advising developers to keep exact values needed for decisions and follow-up calls, summarize relevant bulk information, and re…
Anthropic has listed Growomat, a hosted MCP connector that has run since June, in the Claude connectors directory, replacing the previous setup that required users to copy and paste the URL https://mcp.growomat.com/mcp i…
Amp now lets users choose the models behind its builtin modes and add their own custom agents to the Dial, with two tabs in Settings → Mode Dial for configuration. The Tune Modes tab lets users set the model and reasonin…
A first-person review of the Astra model concludes it is "very good" and "very smart and capable" but limited by the tools, harnesses, and context it can access, with the author noting Astra is increasingly competent at …
Claude Code version 2.1.267 adds MCP servers through the `claude mcp add` command, which writes the server entry to a config file and loads its tools on the next session, according to Anthropic's documentation checked 10…
Parameter AI is promoting autonomous AI agents that continuously pentest code, cloud, and dependencies, arguing that penetration testing cadence matters more than framework choice among PTES, OWASP, NIST SP 800-115, and …
A Digital Applied guide published September 12, 2026 argues that rising AI usage only shows more work is passing through an AI system, not that teams are completing more useful work, and proposes a decision record tying …
A proposed build workflow for coding agents advises developers to construct interactive product demos around a single visitor decision, with explicit labels distinguishing real application logic, sample data, and simulat…
A Digital Applied reference published September 12, 2026 proposes splitting managed AI agent state into three separately owned layers — conversation, execution and business — so that session recovery is not mistaken for …
A new design guide proposes an application pattern for keeping voice agents useful during slow backend tool calls, requiring explicit task states such as pending, running, completed, failed and uncertain, and tracking a …
Kalshi MCP version 1.2.0, a hosted Model Context Protocol server for Kalshi prediction markets, was published to the official MCP registry on 2026-09-09 and filed under Search by the mcpindex index. The index's semantic …