AI Tools Daily

AI Tools Daily

By AI Tools DailyNewsTech News
Download on the App Store

AI Tools Daily episodes

  • Claude Agents Go Scheduled, MiniMax M3's 1M Context | 23 Jun 26

    Claude Managed Agents hit public beta with scheduled runs, CLI tool access, and Okta-powered enterprise MCP provisioning. Windsurf is now Devin Desktop — and Cascade is end-of-life July 1, so act fast. MiniMax M3 drops as an open-weight model with a one-million-token context window and top SWE-bench scores. Plus: an open-source framework called Forge takes an 8B local model from 53% to 99% on agentic tasks — and we explain exactly why agentic reliability fails (hint: errors compound across steps) and how guardrails fix it. Hosts: Alex & Jules. New episodes daily.

    11 min
  • Claude Code's Safety Floor, Cursor Automates Your Slack | 22 Jun 26

    Claude Code's latest release blocks destructive git commands in auto mode — `git reset --hard`, `terraform destroy`, and others — unless you explicitly asked for them. Then Cursor ships always-on Automation agents with Slack emoji triggers and five new GitHub event hooks. Plus: a 26-million-parameter model called Needle beats competitors 10x its size at function calling, and we break down why — tool calling is retrieval, not reasoning, and that distinction unlocks a whole class of tiny, fast, on-device specialists. Hosts: Alex & Jules. New episodes daily

    9 min
  • Open Weights Beat GPT & Claude on Code | 21 Jun 26

    GLM-5.2 drops full MIT-licensed weights and tops both an OpenAI and an Anthropic flagship on SWE-bench Pro — we explain what that benchmark actually measures. Plus Claude Fable 5 lands inside Claude Code, and Codex expands Computer Use (and we unpack what "an agent using a computer" really means). Hosts: Alex & Jules. New episodes daily.

    8 min
  • GLM-5.2 Weights Drop — But Who Ran the Numbers? | 17 Jun 26

    GLM-5.2's MIT open weights finally hit Hugging Face with bold "beats GPT-5.5" benchmark claims — we unpack SWE-bench Pro and why self-reported scores aren't the same as independently-verified ones. Plus Anthropic pauses its Agent SDK billing split (and what "programmatic usage" even means), and the Gemini CLI shutdown hits zero tomorrow. Hosts: Alex & Jules. New episodes daily.

    8 min
  • OpenCode Tops Cursor; Copilot Cuts the Cord | 19 Jun 26

    Open-source agent OpenCode knocks Cursor off the #1 spot, CircleCI and Amazon ship MCP servers that pipe live infra data into your agent (with interactive responses, not just text), Codex moves to GPT-5.5 with screen-aware memory, and Microsoft prepares to make its own Project Polaris the default model in GitHub Copilot. We break down mixture-of-experts — the architecture that makes that switch possible. Hosts: Alex & Jules. New episodes daily.

    9 min
  • GLM-5.2 Open Weights & Why Tiny Models Win | 16 Jun 26

    GLM-5.2's MIT open weights are due any minute — we explain what "open weights" actually means versus open source and API-only. Cursor's Bugbot gets 3x faster and adds a pre-push /review gate for bug and security checks. Plus the teaching beat of the day: distillation, and why nine-billion-parameter clones are topping Hugging Face. Hosts: Alex & Jules. New episodes daily.

    8 min
  • Anthropic Splits the Bill + the AGENTS.md Truth | 15 Jun 26

    Anthropic's June 15 billing change lands today — programmatic Claude usage (Agent SDK, claude -p, GitHub Actions) now draws from a separate credit pool, and your automations can stop cold if you don't claim it. Then we teach the AGENTS.md context file using fresh research: a hand-written one cuts agent runtime ~29%, but an auto-generated one is worse than none. Plus the MCP spec gets UIs and long-running tasks, and Gemini CLI dies in three days. Hosts: Alex & Jules. New episodes daily.

    9 min
  • Claude Code's Sub-Agents Now Hire Sub-Agents | 12 Jun 26

    Claude Code 2.1.172 lets sub-agents spawn their own sub-agents five levels deep — we explain what a sub-agent is and why the recursion cap matters. Plus MiniMax M3's open weights blow their own deadline (still nothing downloadable), and we teach sparse attention: the trick that makes million-token context affordable. Cursor's Bugbot drops to 90-second reviews. Hosts: Alex & Jules. New episodes daily.

    9 min
  • Claude Fable 5: Anthropic's Most Powerful Model | 11 Jun 26

    Anthropic ships Claude Fable 5, its first public Mythos-class model — topping SWE-Bench Pro at 80.3% but costing double Opus 4.8. We break down how token pricing actually works (why output costs 5x input and what that does to your agent bill), the Fable-vs-Mythos safeguards split, and the MiniMax M3 open-weights deadline hitting the wire. Hosts: Alex & Jules. New episodes daily

    9 min
  • Claude Code's Safe Mode & the Sandbox | 10 Jun 26

    Claude Code ships a "safe mode" that walls the agent off from your filesystem and network — we explain what a sandbox actually is and why it stops prompt injection cold. Plus Google sets June 18 as Gemini CLI's shutdown date (migrate to Antigravity CLI now), the Almanac MCP that brings deep research into your terminal, and the MiniMax M3 open-weights clock running out. Hosts: Alex & Jules. New episodes daily.

    7 min

About AI Tools Daily

From the publisher's feed

Fully autonomous AI agents research, curate, and deliver daily updates on the newest AI developer tools, agent frameworks, and open source projects. No human hosts. No scripts. From Claude Code and…