<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Vibecoding.cz — English</title><description>Practical articles on Claude Code, agentic development, and the AI coding market.</description><link>https://www.vibecoding.cz/</link><atom:link href="https://www.vibecoding.cz/en/feed.xml" rel="self" type="application/rss+xml"/><language>en</language><item><title>Swarms: parallel worker agents with a live board, and Auto as the default mode</title><link>https://www.vibecoding.cz/en/antigravity-weekly-2026-w46/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/antigravity-weekly-2026-w46/</guid><description>Swarms let a coordinator agent split a task across worker agents you can watch on a full-screen board. Auto also becomes the default Agent Mode, which changes how headless runs start.</description><pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Fable 5.2 becomes the default Fable model as permission defaults get stricter</title><link>https://www.vibecoding.cz/en/claude-code-weekly-2026-w46/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-weekly-2026-w46/</guid><description>Claude Fable 5.2 is now the default Fable model with a 1M context window. Sandboxed Bash, read-only git and scheduled tasks now ask before they run, and Code Mode arrives as an experiment.</description><pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Side-chat tabs arrive in the terminal workspace</title><link>https://www.vibecoding.cz/en/codex-weekly-2026-w46/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/codex-weekly-2026-w46/</guid><description>The opt-in terminal workspace now supports multiple side-chat tabs and reopens saved local side chats after a restart. The week also brings provider management from the terminal, collapsible transcripts, audio attachments and animated GIFs.</description><pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Search session contents from the /resume picker</title><link>https://www.vibecoding.cz/en/grok-build-weekly-2026-w46/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-weekly-2026-w46/</guid><description>The /resume picker now searches what was said inside your sessions as you type and opens a match directly. The week also brings a minimal sandbox profile, a workspace_write option and new keyboard shortcuts for long messages and copying replies.</description><pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Gemini 4 Hydrogen: The Fast Workhorse</title><link>https://www.vibecoding.cz/en/gemini-4-hydrogen-the-fast-workhorse/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/gemini-4-hydrogen-the-fast-workhorse/</guid><description>Google released it as the successor to the cheap Flash, and it immediately leapfrogged the month-old Gemini 4 Argon. From December, it will cost one fifth of Argon&apos;s price. It&apos;s fast, available everywhere, and you don&apos;t pay extra for long context. That&apos;s the new Hydrogen model.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate></item><item><title>How to Prompt Claude Fable 5.2: The Model That Listens a Little Too Well</title><link>https://www.vibecoding.cz/en/how-to-prompt-claude-fable-5-2-the-model-that-listens-a-little-too-well/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/how-to-prompt-claude-fable-5-2-the-model-that-listens-a-little-too-well/</guid><description>How do you properly prompt the new Claude Fable 5.2 model and get the most out of it? Anthropic has released a dedicated prompting guide. Let&apos;s walk through what&apos;s different!</description><pubDate>Wed, 11 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Your editor can now use agy as its agent over ACP</title><link>https://www.vibecoding.cz/en/antigravity-weekly-2026-w45/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/antigravity-weekly-2026-w45/</guid><description>`agy acp` turns the CLI into an Agent Client Protocol server for editors. The same releases add remote-agent sessions, local projects, stricter permission handling and a changed headless timeout exit code.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Claude Fable 5.2 is the smartest again. But Opus is enough for most of you</title><link>https://www.vibecoding.cz/en/claude-fable-5-2-is-the-smartest-again-but-opus-is-enough-for-most-of-you/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-fable-5-2-is-the-smartest-again-but-opus-is-enough-for-most-of-you/</guid><description>Fable 5.1 cost twice as much as Opus 5.5, and in independent benchmarks it still fell short of it. The new version is fortunately in a different place: six points above Opus, a fifth cheaper, and faster than its predecessor. Even Anthropic itself advises starting with Opus, though. So when does it make sense to pay extra for Fable?</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Agent command center opens from the CLI, and the session picker marks unread results</title><link>https://www.vibecoding.cz/en/codex-weekly-2026-w45/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/codex-weekly-2026-w45/</guid><description>Codex CLI 0.167.0 adds `codex agents --no-daemon` to open the agent command center, with unread markers in the command center and session picker. The release also brings backend switching with kept drafts, opt-in system proxy settings and expandable patches in approval prompts.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate></item><item><title>The @ file picker now reaches outside your workspace, and chips are editable</title><link>https://www.vibecoding.cz/en/grok-build-weekly-2026-w45/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-weekly-2026-w45/</guid><description>Typing an @ path that starts with /, ~/ or ../ now lists files from that folder, and picked files can be edited in place. The week also brings Shift+Enter in more terminals, full Vim editing in the prompt and clearer Auto-review cards.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Muse Ember: Haiku-level performance at a third of the price, but only for routine work</title><link>https://www.vibecoding.cz/en/muse-ember-haiku-level-performance-at-a-third-of-the-price-but-only-for-routine-work/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/muse-ember-haiku-level-performance-at-a-third-of-the-price-but-only-for-routine-work/</guid><description>Meta has released a cheap model that matched Claude Haiku 5.5 in an independent index, with tasks costing a third as much. What should you use it for, what should you test it on, and when should you stick with Haiku, DeepSeek or Luna? I tested it!</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Sessions in Claude Code Can Talk to Each Other</title><link>https://www.vibecoding.cz/en/sessions-in-claude-code-can-talk-to-each-other/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/sessions-in-claude-code-can-talk-to-each-other/</guid><description>Do you have three Claude Code sessions open and keep copying output from one into another? You don&apos;t have to. Sessions can send messages to each other, and Claude can figure out on its own when it should. The feature has been in Claude Code since August, but a lot of people are only learning about it from a viral tutorial that is mostly made up. Let&apos;s go over where and how it actually works.</description><pubDate>Wed, 04 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Paste or drag images straight into the prompt, plus hooks that can stop a turn</title><link>https://www.vibecoding.cz/en/antigravity-weekly-2026-w44/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/antigravity-weekly-2026-w44/</guid><description>Image paste and drag-and-drop now attach images in the prompt, including over SSH. PreInvocation hooks can stop a turn before the model is called, and JSON runs report why they ended early.</description><pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate></item><item><title>ultrathink now runs at xhigh effort, and subagent effort has its own setting</title><link>https://www.vibecoding.cz/en/claude-code-weekly-2026-w44/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-weekly-2026-w44/</guid><description>ultrathink in the terminal and VS Code now runs its turn at xhigh effort, and CLAUDE_CODE_SUBAGENT_EFFORT sets what subagents run at. The week also brings a write guard for paths outside your working directories, idle compaction in VS Code and the Desktop Code tab, and the new Ultra plan.</description><pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Google Cloud Vertex AI arrives as a built-in provider</title><link>https://www.vibecoding.cz/en/codex-weekly-2026-w44/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/codex-weekly-2026-w44/</guid><description>Codex CLI 0.165.0 and 0.166.0 add a built-in Google Cloud Vertex AI provider, switched on with model_provider = &apos;gcp&apos;, plus GPT-6.1 Terra in the Vertex AI and Bedrock catalogs. Memory v2, mid-conversation reasoning changes and keyboard transcript selection also land.</description><pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Experimental vim editing in the prompt, plus fail-closed hooks</title><link>https://www.vibecoding.cz/en/grok-build-weekly-2026-w44/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-weekly-2026-w44/</guid><description>The prompt gets experimental vim modes, hooks can now deny instead of failing open, and subagent output in the transcript becomes clickable. Three releases, 1.0.71 to 1.0.73, cover October 26 to 28, 2026.</description><pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate></item><item><title>How Claude Code Handles Memory About Your Project</title><link>https://www.vibecoding.cz/en/how-claude-code-handles-memory-about-your-project/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/how-claude-code-handles-memory-about-your-project/</guid><description>You correct Claude on the same thing for the second time and wonder whether it will remember it this time. Claude Code now has two layers of persistent memory: rules that you write, and notes that it keeps on its own. Both are plain Markdown files on your disk. That is the biggest advantage of the whole approach and also its weakness, because a note in a file only helps if Claude remembers to open it. How do you work with this so that you end up satisfied?</description><pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Gemini 4 Carbon has caught up with Opus, but you can&apos;t use it yet</title><link>https://www.vibecoding.cz/en/gemini-4-carbon-has-caught-up-with-opus-but-you-cant-use-it-yet/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/gemini-4-carbon-has-caught-up-with-opus-but-you-cant-use-it-yet/</guid><description>After a year, Google is back at the top of the independent Artificial Analysis leaderboard, and on the Vals Index it matches Claude Opus 5.5 at a third of the price. If you want to try it right now, you&apos;re out of luck: for the time being, Carbon goes only to vetted defenders in cybersecurity. Where does it actually fit in Google&apos;s lineup, and what should you do until it becomes available?</description><pubDate>Thu, 29 Oct 2026 00:00:00 GMT</pubDate></item><item><title>GPT-6.1 Terra only pays off at half throttle</title><link>https://www.vibecoding.cz/en/gpt-6-1-terra-only-pays-off-at-half-throttle/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/gpt-6-1-terra-only-pays-off-at-half-throttle/</guid><description>Ten times pricier per token than Claude Haiku 5.5 and four times pricier than DeepSeek V4.1 Flash. Still, at medium effort Terra delivers more than both rivals at maximum, for less than half the cost per task. The catch is how you run it. From xhigh upward, the same company&apos;s GPT-6.1 Sol is both better and cheaper.</description><pubDate>Wed, 28 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Search arrives in /diff and the artifact viewer, with n/N stepping through every match</title><link>https://www.vibecoding.cz/en/antigravity-weekly-2026-w43/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/antigravity-weekly-2026-w43/</guid><description>Version 1.4.3 adds `/` search to `/diff` and match-by-match `n`/`N` in the artifact viewer. The rest of the week brings diff folding, Auto mode on Windows 11 24H2+ and a `Configure MCP servers...` row in `/plugin`.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Code Review gets review levels, REVIEW.md includes and a stop command</title><link>https://www.vibecoding.cz/en/claude-code-weekly-2026-w43/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-weekly-2026-w43/</guid><description>Code Review now has Quick, Standard and Deep levels, REVIEW.md can include other files, and `@claude review stop` ends an always-on subscription. The week also brings model and effort controls in agent view, gateway deny-lists and sandbox changes.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Reusable prompts in ten numbered /presets slots</title><link>https://www.vibecoding.cz/en/codex-weekly-2026-w43/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/codex-weekly-2026-w43/</guid><description>Codex CLI 0.164.0 adds ten numbered /presets slots for reusable prompts and slash commands. It also adds inline arguments for /model and /permissions, image previews in the fullscreen transcript, and a default-on Amazon Bedrock setup wizard.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Fewer Subagents, or When Too Many Agents Hurt in Claude...</title><link>https://www.vibecoding.cz/en/fewer-subagents-or-when-too-many-agents-hurt-in-claude/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/fewer-subagents-or-when-too-many-agents-hurt-in-claude/</guid><description>It&apos;s tempting. You split an agent into specialists, each team takes its own piece, and an orchestrator ties it all together at the top. Finally, dozens of agents are working for you! It looks clean, it grows along with the organization, and everyone knows what they&apos;re responsible for. Based on its experience with enterprise deployments, Anthropic claims that such a setup often performs worse, runs slower, and costs more than a single agent with Skills. So when should you merge subagents, when should you keep them, and how do you tell that merging didn&apos;t break something?</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Steer a running turn or send now: new follow-up controls in the prompt</title><link>https://www.vibecoding.cz/en/grok-build-weekly-2026-w43/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-weekly-2026-w43/</guid><description>A new follow_up_behavior setting lets Enter steer a running turn at its next step, and Ctrl+Enter or Ctrl+I cancels and sends. The week also brings cross-folder /resume search, managed-only hooks in untrusted folders and worktrees that inherit approvals.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Built-in skills for 3D pages and browser games that test their own output</title><link>https://www.vibecoding.cz/en/muse-code-weekly-2026-w43/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/muse-code-weekly-2026-w43/</guid><description>Three new built-in skills build 3D pages and browser games and play them before handoff, while /game-3d-visuals gets faster. The week also brings local port rules for sandboxed commands, enforced deny rules in :ask-me profiles and a muse usage command.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Kimi K3.1 Pro: the best open model is slow and not as cheap as it looks</title><link>https://www.vibecoding.cz/en/kimi-k3-1-pro-the-best-open-model-is-slow-and-not-as-cheap-as-it-looks/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/kimi-k3-1-pro-the-best-open-model-is-slow-and-not-as-cheap-as-it-looks/</guid><description>Three points behind Claude Opus 5.5, downloadable weights, and half the price per token. On paper, it&apos;s the best open model ever released. In practice, though, it responds slowly and thinks a lot. Compared to Opus, depending on the task and settings, you save two thirds or nothing at all, and compared to GPT-6.1 Sol or DeepSeek it&apos;s expensive. So who is it actually worth it for?</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Claude Duet: The Assistant That Reaches Out on Its Own. Once Every Few Hours</title><link>https://www.vibecoding.cz/en/claude-duet-the-assistant-that-reaches-out-on-its-own-once-every-few-hours/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-duet-the-assistant-that-reaches-out-on-its-own-once-every-few-hours/</guid><description>It&apos;s tempting, I&apos;ll admit. In the morning your assistant tells you what&apos;s on fire, during the day it watches your email, your calendar and your unpaid invoices, and it only pings you when it needs a decision from you. Anthropic has now released exactly that kind of agent (via a waitlist) into a field of several souped-up competitors. What does it promise? And what does it actually do?</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>The First Exploit of the AI Agent Era</title><link>https://www.vibecoding.cz/en/the-first-exploit-of-the-ai-agent-era/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/the-first-exploit-of-the-ai-agent-era/</guid><description>For twenty years, only a handful of people in the world could write a working exploit for a memory bug. Last Wednesday that stopped being true. Two vulnerabilities in the PHP core sat in the code for more than a decade, then a language model chained them together in a week. And now the finished script is circulating on the internet.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Add-ons that promise to make Claude Code cheaper: I measured them on 84 runs</title><link>https://www.vibecoding.cz/en/add-ons-that-promise-cheaper-claude-code-measured-on-84-runs/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/add-ons-that-promise-cheaper-claude-code-measured-on-84-runs/</guid><description>Authors of Claude Code add-ons promise savings from 20 to 98%. I ran the five most recommended ones (Caveman, Ponytail, RTK, Context Mode and Headroom) against plain Claude Code on a real project. Only one saved money consistently, and a shorter CLAUDE.md helped more than any add-on.</description><pubDate>Tue, 20 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Auto mode approves routine actions and still stops before irreversible ones</title><link>https://www.vibecoding.cz/en/antigravity-weekly-2026-w42/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/antigravity-weekly-2026-w42/</guid><description>Auto mode cuts approval prompts for file edits and safe commands while keeping a check on irreversible actions and sandbox escapes. The week also brings /compact, a folding /diff view and terminal status reporting.</description><pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Idle compaction keeps long conversations cheap after a break</title><link>https://www.vibecoding.cz/en/claude-code-weekly-2026-w42/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-weekly-2026-w42/</guid><description>Claude Code now compacts a 200K+ token terminal conversation after about 54 idle minutes, before the one-hour prompt cache expires. The week also tightens MCP allow rules and aligns subagent defaults across the terminal, SDK, desktop and VS Code.</description><pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Search and change settings from /config without leaving the session</title><link>https://www.vibecoding.cz/en/codex-weekly-2026-w42/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/codex-weekly-2026-w42/</guid><description>Codex CLI 0.163.0 lets you search preferences across categories in /config and shows which changes need a restart. It also adds mid-response steering, a configurable leader key in /keymap and parallel read-only tools.</description><pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Resume sessions by id or title with /resume</title><link>https://www.vibecoding.cz/en/grok-build-weekly-2026-w42/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-weekly-2026-w42/</guid><description>/resume and grok --resume now find a session by id or title, with exact matches opening directly. The week also brings copy buttons for code blocks, an args list for command hooks, and a breaking change to memory.</description><pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate></item><item><title>The agent can now generate and edit images, and unattended runs get tighter controls</title><link>https://www.vibecoding.cz/en/muse-code-weekly-2026-w42/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/muse-code-weekly-2026-w42/</guid><description>Muse Code&apos;s agent can generate and edit images when your model catalog has an image model, and /blender rebuilds reference images as Blender scenes. Launch flags for unattended runs and a stricter permission-profile refusal also landed.</description><pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate></item><item><title>When Harari Talks About AI After a Booed Macinka</title><link>https://www.vibecoding.cz/en/when-harari-talks-about-ai-after-a-booed-macinka/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/when-harari-talks-about-ai-after-a-booed-macinka/</guid><description>At a surreal conference hosted by a billionaire&apos;s wife, Foreign Minister Petr Macinka spoke about migration first, and then the sociohistorian Yuval Noah Harari talked about AI as a form of migration. The contrast could not have been greater.</description><pubDate>Tue, 13 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Plugin marketplace arrives with /plugin, and defaults get stricter</title><link>https://www.vibecoding.cz/en/antigravity-weekly-2026-w41/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/antigravity-weekly-2026-w41/</guid><description>The new /plugin command installs bundles of skills, MCP servers, subagents, rules, and hooks from the official marketplace. The same week brings a medium Verbosity default and tighter permission and sandbox behavior.</description><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Haiku 5.5 becomes the default Haiku, and sub-agents get their own effort and compaction window</title><link>https://www.vibecoding.cz/en/claude-code-weekly-2026-w41/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-weekly-2026-w41/</guid><description>Claude Haiku 5.5 is now the default Haiku model on the Anthropic API, with a 1M context window. The week also brings sub-agent effort and compaction controls, fail-closed hooks, and a new default MCP protocol version.</description><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate></item><item><title>GPT-6.1 Sol becomes the default model, and Bedrock gains Ultra reasoning</title><link>https://www.vibecoding.cz/en/codex-weekly-2026-w41/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/codex-weekly-2026-w41/</guid><description>GPT-6.1 Sol is the new default model in the bundled and Amazon Bedrock catalogs. The week also brings /mcp login, opt-in Daybreak, managed Git worktree tools and TUI transcript copy.</description><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Transcript density and timestamps: control how much of a long session you see</title><link>https://www.vibecoding.cz/en/grok-build-weekly-2026-w41/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-weekly-2026-w41/</guid><description>Grok Build 1.0.50 to 1.0.52 adds /density and /timestamps for tuning transcript detail, plus /workflow for background runs and stricter safeguards around worktree deletion.</description><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Plugin management is on by default, plus per-run config flags and safer MCP configs</title><link>https://www.vibecoding.cz/en/muse-code-weekly-2026-w41/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/muse-code-weekly-2026-w41/</guid><description>Muse Code 1.4.3 to 1.4.6 turns plugin management on by default and adds repeatable -c config flags, env-expanded MCP configs and corporate CA support. Session handling, two new hooks and muse serve permission defaults round out the week.</description><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Decision Models: Cheap AI That Picks the Right Answer</title><link>https://www.vibecoding.cz/en/decision-models-cheap-ai-that-picks-the-right-answer/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/decision-models-cheap-ai-that-picks-the-right-answer/</guid><description>Did you pull the right data out of a document? Did you correctly detect the OK button in an interface? Decision models handle this better than a regular LLM. Jev started it, and within two weeks more than ten models appeared around this problem. Cloudflare, Perplexity, Amazon and OpenAI have all entered the game. I&apos;ll show you how to find your way around them, what they cost and, above all, which tasks they&apos;re really good for.</description><pubDate>Mon, 05 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Mods for Claude Code: Anthropic Lets You Under the Hood</title><link>https://www.vibecoding.cz/en/mods-for-claude-code-anthropic-lets-you-under-the-hood/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/mods-for-claude-code-anthropic-lets-you-under-the-hood/</guid><description>What has always annoyed me about Claude Code is how it prints colorful diffs while it works, and I don&apos;t want to read them. But there was no way to hide them. Until now: Anthropic has introduced Mods, which let you properly rebuild the interface.</description><pubDate>Fri, 02 Oct 2026 00:00:00 GMT</pubDate></item><item><title>OpenAI DevDay 2026: Dots Personal Agents and Other News</title><link>https://www.vibecoding.cz/en/openai-devday-2026-dots-personal-agents-and-other-news/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/openai-devday-2026-dots-personal-agents-and-other-news/</guid><description>At DevDay 2026, OpenAI announced more than twenty new things. For us vibecoders, there are three main themes: Dots as agents that work continuously, an upgraded Codex as an environment for delegated development, and an expanded platform for custom apps and automations. I won&apos;t leave the GPT-6.1 model aside either. Let&apos;s go through it all.</description><pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate></item><item><title>The $200 Codex Pro Plan Is Back. With Half the Volume</title><link>https://www.vibecoding.cz/en/the-200-codex-pro-plan-is-back-with-half-the-volume/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/the-200-codex-pro-plan-is-back-with-half-the-volume/</guid><description>Not long ago, OpenAI pulled its $200 plan because of high demand. Now it&apos;s bringing it back, reworked. How much work does a $200-a-month subscription actually buy you?</description><pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Who profits when you run an LLM at home</title><link>https://www.vibecoding.cz/en/who-profits-when-you-run-an-llm-at-home/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/who-profits-when-you-run-an-llm-at-home/</guid><description>Lately there&apos;s a lot going on around locally run LLMs. Performance keeps rising, and with it the hope that we can get rid of subscriptions to closed models. But things are boiling under the surface. Why would NVIDIA buy Hugging Face for $13 billion, when its most famous piece of software, llama.cpp, runs AI on hardware NVIDIA doesn&apos;t sell? Let&apos;s take a look at this world.</description><pubDate>Mon, 21 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Jev: the model that doesn&apos;t write, it just decides. How to use it and when it pays off</title><link>https://www.vibecoding.cz/en/jev-the-model-that-doesnt-write-it-just-decides-how-to-use-it-and-when-it-pays-off/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/jev-the-model-that-doesnt-write-it-just-decides-how-to-use-it-and-when-it-pays-off/</guid><description>It has the AI community buzzing. Jev from TypeSafe can&apos;t write a single sentence, but for a fraction of the price of Haiku it answers a yes/no question, picks from a list of options or gives a score. So I went to try it out...</description><pubDate>Sat, 19 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Claude Code Projects Now Split Up the Work on Their Own</title><link>https://www.vibecoding.cz/en/claude-code-projects-now-split-up-the-work-on-their-own/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-projects-now-split-up-the-work-on-their-own/</guid><description>Anthropic has just launched redesigned Projects in Claude Code. A project is now a single long-running conversation in which Claude acts as a coordinator. It breaks the assignment into individual threads, each run as a separate session on its own branch, and they report their results back.</description><pubDate>Fri, 18 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Černá AI Festival Lit Up Ostrava</title><link>https://www.vibecoding.cz/en/cerna-ai-festival-lit-up-ostrava/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/cerna-ai-festival-lit-up-ostrava/</guid><description>A beautifully designed event, both visually and in its programming, set in the striking surroundings of the former Vítkovice ironworks. It was well worth attending, and it raised the bar for events in the Czech Republic. So what was it about?</description><pubDate>Wed, 16 Sep 2026 00:00:00 GMT</pubDate></item><item><title>How to Tell a Good AI Workshop from a Bad One</title><link>https://www.vibecoding.cz/en/how-to-tell-a-good-ai-workshop-from-a-bad-one/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/how-to-tell-a-good-ai-workshop-from-a-bad-one/</guid><description>The debate about the quality of the various AI trainings, and especially of the trainers themselves, has resurfaced lately. A few experiences and a view from the inside might be useful here.</description><pubDate>Mon, 14 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Cursor Projects Shows a Great New Approach to Managing Agents</title><link>https://www.vibecoding.cz/en/cursor-projects-shows-a-great-new-approach-to-managing-agents/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/cursor-projects-shows-a-great-new-approach-to-managing-agents/</guid><description>How do you make AI developer tools better? How do you fix the way they handle context so they don&apos;t slow you down as a human, and by extension don&apos;t slow down the AI agents either? Cursor Projects is trying to answer that. I can&apos;t wait for the others to copy it!
</description><pubDate>Sun, 13 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Anthropic&apos;s head of AI safety testing: We don&apos;t have a plan yet for aligning superintelligence</title><link>https://www.vibecoding.cz/en/anthropics-head-of-ai-safety-testing-we-dont-have-a-plan-yet-for-aligning/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/anthropics-head-of-ai-safety-testing-we-dont-have-a-plan-yet-for-aligning/</guid><description>At the beginning of September, Jacob Coxon announced he was leaving Anthropic. The name probably won&apos;t mean anything to you: he did pre-training research on models, earlier at OpenAI, most recently at Anthropic. He accompanied his resignation with a thread on X that took off immediately. In it, he claims that both OpenAI and Anthropic are heading toward self-improving superintelligence in a way he doesn&apos;t consider responsible.</description><pubDate>Wed, 09 Sep 2026 00:00:00 GMT</pubDate></item><item><title>OpenClaw 2.0 Does Multiplayer. Watch Out for It!</title><link>https://www.vibecoding.cz/en/openclaw-2-0-does-multiplayer-watch-out-for-it/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/openclaw-2-0-does-multiplayer-watch-out-for-it/</guid><description>It&apos;s tempting. One agent, one machine, a whole team in a single thread, with work handed off along with its context. The new version of OpenClaw lures you in with support for multiple users in one installation, in one thread. Limits? There are some!</description><pubDate>Mon, 31 Aug 2026 00:00:00 GMT</pubDate></item><item><title>The Case of OpenAI&apos;s Rogue AI Agents Attacking Hugging Face</title><link>https://www.vibecoding.cz/en/the-case-of-openais-rogue-ai-agents-attacking-hugging-face/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/the-case-of-openais-rogue-ai-agents-attacking-hugging-face/</guid><description>A month ago, a widely reported case saw OpenAI&apos;s AI agents attack Hugging Face servers. Now it has a resolution. The company itself published a technical report, and an independent investigation has come out as well. The two stories diverge somewhat, and they are more alarming than they first appeared.</description><pubDate>Fri, 28 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Hardware for Local AI: What to Buy in August 2026, and Whether It Pays Off</title><link>https://www.vibecoding.cz/en/hardware-for-local-ai-what-to-buy-in-august-2026-and-whether-it-pays-off/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/hardware-for-local-ai-what-to-buy-in-august-2026-and-whether-it-pays-off/</guid><description>It&apos;s incredibly tempting. You buy an AI computer and run local models for the cost of a hardware installment plan. For this game to add up, the hardware has to be available, open models have to be strong enough, and the prices of the closed ones can&apos;t collapse. So what should you pick?</description><pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate></item><item><title>An AI watermark proves touch, not authorship. What should we do about labeling AI content?</title><link>https://www.vibecoding.cz/en/an-ai-watermark-proves-touch-not-authorship-what-should-we-do-about-labeling-ai-content/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/an-ai-watermark-proves-touch-not-authorship-what-should-we-do-about-labeling-ai-content/</guid><description>On Monday, Anthropic confirmed that it embeds an invisible watermark in text generated by Claude. The hate toward the EU started immediately, blaming it for the move, but the EU requires nothing of the sort. Where are the limits of AI in text, and how do I use AI to write my own?</description><pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate></item><item><title>The Agent Deleted Someone Else&apos;s Booking: What AI Agents Do to a Sloppy SaaS</title><link>https://www.vibecoding.cz/en/the-agent-deleted-someone-elses-booking-what-ai-agents-do-to-a-sloppy-saas/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/the-agent-deleted-someone-elses-booking-what-ai-agents-do-to-a-sloppy-saas/</guid><description>An Australian user asked OpenClaw to book a gym class. The agent found a missing ownership check in the booking API and canceled a stranger&apos;s reservation. A breakdown of what happened technically, and what it means for anyone running a SaaS.</description><pubDate>Tue, 11 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Jeff Dean Is Leaving Google. He Has Already Become a Legend.</title><link>https://www.vibecoding.cz/en/jeff-dean-is-leaving-google-he-has-already-become-a-legend/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/jeff-dean-is-leaving-google-he-has-already-become-a-legend/</guid><description>After 27 years, Google&apos;s most senior technical person is leaving. He&apos;s taking three other stars with him and founding Discovery Loop, and Alphabet has chipped in funding and is leasing him machines. The market still wiped $185 billion off Alphabet&apos;s value.</description><pubDate>Mon, 10 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Grok Build Reaches Version 1.0</title><link>https://www.vibecoding.cz/en/grok-build-reaches-version-1-0/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-reaches-version-1-0/</guid><description>SpaceXAI has released the first stable version of Grok Build, its terminal-based coding agent. The changelog has thirty items, and every one of them is a fix. What has actually changed in the tool over two months, and what does the 1.0 label mean?</description><pubDate>Fri, 07 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Meta Muse Code: When Zuckerberg Teaches AI to Code</title><link>https://www.vibecoding.cz/en/meta-muse-code-when-zuckerberg-teaches-ai-to-code/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/meta-muse-code-when-zuckerberg-teaches-ai-to-code/</guid><description>Meta has released Muse Code, its first coding agent, along with the Muse Spark 1.2 model. It&apos;s a terminal-based beta aimed at large repositories. Two ideas inside are worth your attention, the pricing is aggressive, and the benchmarks can&apos;t be trusted.</description><pubDate>Thu, 06 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Claude Code rewrote /fork. For the second time this month. And this time, it was useful</title><link>https://www.vibecoding.cz/en/claude-code-rewrote-fork-for-the-second-time-this-month-and-this-time-it-was-useful/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-rewrote-fork-for-the-second-time-this-month-and-this-time-it-was-useful/</guid><description>Anthropic has changed the /fork in Claude Code for the second time this month; it’s now a background agent with inherited context. This time, it’ll come in handy. How do you work with it?</description><pubDate>Wed, 03 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Dynamic Workflows in Claude Code: An Announced Experiment Now Set to Be Put to the Test</title><link>https://www.vibecoding.cz/en/dynamic-workflows-in-claude-code/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/dynamic-workflows-in-claude-code/</guid><description>Anthropic has introduced a new feature called Dynamic Workflows in Claude Code, alongside the Opus 4.8 model. Claude writes its own orchestration script, which launches tens to hundreds of parallel subagents in a single session, and checks the results itself before delivering them to you.</description><pubDate>Sun, 31 May 2026 00:00:00 GMT</pubDate></item><item><title>Grok Build CLI – xAI Enters the Terminal Coding Agent Market</title><link>https://www.vibecoding.cz/en/grok-build-cli-x-ai-enters-the-terminal-coding-agent-market/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/grok-build-cli-x-ai-enters-the-terminal-coding-agent-market/</guid><description>xAI has launched Grok Build, a terminal agent for programming, as a direct competitor to Claude Code. For now, it’s only available to SuperGrok Heavy plan subscribers, but a head-to-head battle with Anthropic is on the horizon...</description><pubDate>Sat, 16 May 2026 00:00:00 GMT</pubDate></item><item><title>Claude Code Agent view—it’s a whole new way of working</title><link>https://www.vibecoding.cz/en/claude-code-agent-view-it-s-a-whole-new-way-of-working/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-agent-view-it-s-a-whole-new-way-of-working/</guid><description>Anthropic is striving to demonstrate that swarms of agents are right at home in the Claude Code. To that end, it has introduced Agent View, which allows users to monitor the status of all agents on a single screen. However, this significantly changes the way we work. So what’s new and different?</description><pubDate>Thu, 14 May 2026 00:00:00 GMT</pubDate></item><item><title>Claude Code also offered /goal for automatically reaching a destination</title><link>https://www.vibecoding.cz/en/claude-code-also-offered-goal-for-automatically-reaching-a-destination/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/claude-code-also-offered-goal-for-automatically-reaching-a-destination/</guid><description>Anthropic has added the `/goal` command to Claude Code, which keeps running a task until it’s finished. Typically, it keeps correcting an error until it’s fixed. It’s like a Ralph Wiggum Loop for adults...</description><pubDate>Tue, 12 May 2026 00:00:00 GMT</pubDate></item><item><title>🐑 Sheep Marketplace — how Tomas Jirikovsky and the Czech Minister of Justice came to Bitcoin</title><link>https://www.vibecoding.cz/en/sheep-marketplace-how-tomas-jirikovsky-and-the-czech-minister-of-justice-came-to-bitcoin/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/sheep-marketplace-how-tomas-jirikovsky-and-the-czech-minister-of-justice-came-to-bitcoin/</guid><description>If you don’t live in the Czech Republic, you may not understand this story. A few years ago, when the government wanted to promote tourism in the Czech Republic, it came up with the slogan “land of stories.”</description><pubDate>Mon, 02 Jun 2025 00:00:00 GMT</pubDate></item><item><title>Claude Code Hooks — The Complete Guide</title><link>https://www.vibecoding.cz/en/2025-05-14-claude-code-hooks-complete-guide/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/2025-05-14-claude-code-hooks-complete-guide/</guid><description>Hooks let you run shell commands before and after every Claude Code action. Here&apos;s how to use them to enforce linting, automate git, and build custom workflows that run silently in the background.</description><pubDate>Wed, 14 May 2025 00:00:00 GMT</pubDate></item><item><title>When AI comes for those with political power…</title><link>https://www.vibecoding.cz/en/when-ai-comes-for-those-with-political-power/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/when-ai-comes-for-those-with-political-power/</guid><description>In Western society, the phenomenon is described by Thomas Piketty and Amory Gethin, of the Paris School of economics, when they document the gradual disappearance of political representation of the interests of the working and lower-middle classes in stark contrast to the number of members of that class in society.</description><pubDate>Thu, 07 Mar 2024 00:00:00 GMT</pubDate></item><item><title>The hope of the year: the RISC-V processors</title><link>https://www.vibecoding.cz/en/the-hope-of-the-year-the-risc-v-processors/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/the-hope-of-the-year-the-risc-v-processors/</guid><description>It is already clear that the technology of 2023 will be artificial intelligence, AI.</description><pubDate>Fri, 13 Oct 2023 00:00:00 GMT</pubDate></item><item><title>First Czechoslovak radiotelephone networks: AMR</title><link>https://www.vibecoding.cz/en/first-czechoslovak-radiotelephone-networks-amr/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/first-czechoslovak-radiotelephone-networks-amr/</guid><description>Czechoslovakia did not participate in the development of international mobile networks in the 1970s, the isolation caused by the iron curtain and Russia’s policy towards non-socialist countries was a clear separation of socialist countries from the problems of Western mobile networks.</description><pubDate>Mon, 10 Jul 2017 00:00:00 GMT</pubDate></item><item><title>Strategy of motivation to change the behavior in the wearable and Internet of Things world —…</title><link>https://www.vibecoding.cz/en/strategy-of-motivation-to-change-the-behavior-in-the-wearable-and-internet-of-things-world/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/strategy-of-motivation-to-change-the-behavior-in-the-wearable-and-internet-of-things-world/</guid><description>Why does the best part of wearable electronics and devices using Internet of Things exist? The reason could be to make access to limited resources more efficent by motivating us to change our behavior.</description><pubDate>Wed, 11 May 2016 00:00:00 GMT</pubDate></item><item><title>The fairy tale about a single standard for the Internet of Things</title><link>https://www.vibecoding.cz/en/the-fairy-tale-about-a-single-standard-for-the-internet-of-things/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/the-fairy-tale-about-a-single-standard-for-the-internet-of-things/</guid><description>Standards are good. They allow the prices of products to go down quickly; they eliminate the fear of buying something that will not be compatible soon and they lower the price of service.</description><pubDate>Wed, 10 Feb 2016 00:00:00 GMT</pubDate></item><item><title>Edge Computing and Fog Computing; or, on the Benefits of Fogging</title><link>https://www.vibecoding.cz/en/edge-computing-and-fog-computing-or-on-the-benefits-of-fogging/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/edge-computing-and-fog-computing-or-on-the-benefits-of-fogging/</guid><description>These days, many ways of speeding up the operation of wireless networks arose.</description><pubDate>Thu, 21 Jan 2016 00:00:00 GMT</pubDate></item><item><title>WTF of Energy Efficiency: A Quest for the Right Metrics</title><link>https://www.vibecoding.cz/en/wtf-of-energy-efficiency-a-quest-for-the-right-metrics/</link><guid isPermaLink="true">https://www.vibecoding.cz/en/wtf-of-energy-efficiency-a-quest-for-the-right-metrics/</guid><description>What is the most common issue we solve at Energomonitor, where we develop Internet of Things solution for energy savings?</description><pubDate>Mon, 04 Jan 2016 00:00:00 GMT</pubDate></item></channel></rss>