AI News, 14 April 2026

Today’s issue tracks four terminal-based AI coding tools that all pushed new version numbers within days of each other — Claude Code, Gemini CLI, OpenClaw, and Codex — though none of the fetched release pages include visible changelog text. It also notes an OpenAI customer story about CyberAgent, whose actual body text is missing from the source.
Today’s Stories
Section titled “Today’s Stories”1. Claude Code: v2.1.105
Section titled “1. Claude Code: v2.1.105”What changed: Claude Code published the update on 12 Apr 2026. Anthropic tagged Claude Code version 2.1.105 on GitHub, but the fetched release page shows only the site’s standard navigation menu, not the actual release notes.
Short source excerpt: Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster…
Why it matters: Anyone running Claude Code as a coding agent has a new build available, but without visible notes there is no way to tell whether it changes behavior, fixes bugs, or affects existing workflows.
How to verify it: Open Claude Code’s own release history on GitHub and read the full notes for 2.1.105 before you upgrade a machine you rely on.
Uncertainty: The source does not show which bugs were fixed, what features changed, or whether the update requires any manual step to take effect.
Source: v2.1.105
2. Gemini CLI: Release v0.37.2
Section titled “2. Gemini CLI: Release v0.37.2”What changed: Gemini CLI published the update on 12 Apr 2026. Google pushed a new build, v0.37.2, of the open source Gemini CLI terminal tool to GitHub, but the retrieved page contains no release text, only the site’s fixed menu items.
Short source excerpt: An open-source AI agent that brings the power of Gemini directly into your terminal. - Release Release v0.37.2 · google-gemini/gemini-cli…
Why it matters: People running Gemini CLI as a terminal agent can pull the new build, but with no visible notes it is unclear whether commands, permissions, or existing behavior are affected.
How to verify it: Look up v0.37.2 in Gemini CLI’s own release history on GitHub and read the full change list before rolling it out anywhere that matters.
Uncertainty: There is no indication whether v0.37.2 is a bug fix, a security patch, or a feature release, and nothing about whether older versions stop working.
Source: Release v0.37.2
3. OpenClaw: openclaw 2026.4.14-beta.1
Section titled “3. OpenClaw: openclaw 2026.4.14-beta.1”What changed: OpenClaw published the update on 13 Apr 2026. The OpenClaw project tagged a beta build, 2026.4.14-beta.1, of its personal AI assistant, but the fetched material does not describe what the beta actually changes.
Short source excerpt: Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞 - Release openclaw 2026.4.14-beta.1 · openclaw/openclaw Release…
Why it matters: Because this carries a beta tag, anyone experimenting with OpenClaw as a personal assistant should expect possible instability, though without release notes it is impossible to say which parts of the tool are affected.
How to verify it: Read the beta release’s full notes in OpenClaw’s own GitHub project before installing it on a machine you depend on.
Uncertainty: It is unclear whether the beta label covers a single feature or a larger rebuild, and there is no information on when or whether it becomes a stable release.
Source: openclaw 2026.4.14-beta.1
4. OpenAI Codex: 0.120.0
Section titled “4. OpenAI Codex: 0.120.0”What changed: OpenAI Codex published the update on 10 Apr 2026. Codex, OpenAI’s lightweight terminal coding agent, picked up version number 0.120.0, but the fetched release page shows only GitHub’s standard navigation, no actual change text.
Short source excerpt: Lightweight coding agent that runs in your terminal - Release 0.120.0 · openai/codex Release 0.120.0 · openai/codex · GitHub /…
Why it matters: Anyone using Codex as a terminal coding agent has a new version available, but without visible notes there is no way to judge whether commands, permissions, or pricing changed.
How to verify it: Compare version 0.120.0 against Codex’s own changelog on GitHub and against the version you currently have installed before updating.
Uncertainty: The release page gives no indication of which bugs were fixed, which models are supported, or whether default behavior changed.
Source: 0.120.0
5. OpenAI: CyberAgent moves faster with ChatGPT Enterprise and Codex
Section titled “5. OpenAI: CyberAgent moves faster with ChatGPT Enterprise and Codex”What changed: OpenAI published the update on 8 Apr 2026. OpenAI published a customer story about CyberAgent reportedly working faster with ChatGPT Enterprise and Codex, but the article body is missing from the fetched material, leaving no concrete details to report.
Why it matters: Without an actual story, there is no specific consequence to draw for readers running ChatGPT Enterprise or Codex in smaller setups — only that OpenAI is using the case in its own marketing.
How to verify it: Find the customer story directly on OpenAI’s own site to see what figures or workflows actually back the headline.
Uncertainty: Nothing in the material indicates which features CyberAgent uses, what time savings are claimed, or whether the results would transfer to a smaller setup.
Source: CyberAgent moves faster with ChatGPT Enterprise and Codex
Why It Matters
Section titled “Why It Matters”- The focus is changes that can affect AI products, models, browser experiences, agents, API usage, privacy, pricing, or security.
- A new feature is useful in the home only after access, logging, and rollback have been tested.
- The article separates documented information from editorial judgment and open questions.
Sources and Further Reading
Section titled “Sources and Further Reading”| Publisher | Source | Date |
|---|---|---|
| Claude Code | v2.1.105 | 12 Apr 2026 |
| Gemini CLI | Release v0.37.2 | 12 Apr 2026 |
| OpenClaw | openclaw 2026.4.14-beta.1 | 13 Apr 2026 |
| OpenAI Codex | 0.120.0 | 10 Apr 2026 |
| OpenAI | CyberAgent moves faster with ChatGPT Enterprise and Codex | 8 Apr 2026 |
Editorial Note
Section titled “Editorial Note”SmartBolig.net used automated research, deduplication, and drafting for this issue. Editorial rules require a primary source, verifiable links, and explicit uncertainty.
Comments
SmartBolig // AI Core
SmartBolig AI
Get help with troubleshooting, architecture and concrete configuration. I use broad AI knowledge and find SmartBolig guides when they improve the answer.
Start with a useful question