Tech • AI • Robotics • Game
Grok Bot launches cloud routines as Agentic links Claude to 100s of APIs
Briefing
Grok Bot pushes cloud automation
xAI is positioning Grok Bot as a cloud-run automation layer rather than a conventional chat assistant. Users define bots as persistent roles with instructions, plugins, saved skills and handoffs, then let them execute multi-step work on xAI infrastructure. The desktop app on Windows, macOS and iOS acts mainly as a control surface for setup, monitoring and review. That architecture means scheduled jobs can continue even when the local device is offline or shut down.
Scheduled routines become the key battleground
A central feature of Grok Bot is its routines system, which turns prompts into recurring operations with triggers and schedules. The model is closer to assigning a digital employee a standing brief than opening a one-off chat, with examples like a Weekly Intelligence Analyst that researches topics, writes concise summaries and emails results. The product also supports bot-to-bot delegation, suggesting a push toward modular agent teams. The broader implication is that value is shifting from text generation alone to durable, unattended execution.
Agentic gives Claude tool access
Agentic is pitching itself as a single connection layer that lets an assistant tap hundreds of external services through one setup. In the demonstrated workflow, Claude Desktop is linked through MCP, exposing a callable tool inside the chat rather than forcing separate bespoke integrations. The company’s argument is that models become meaningfully more useful when they can act, fetch live data and hit specialized APIs on demand. That framing places tool access, not just model quality, at the center of the next agent wave.
Live APIs challenge web-search answers
Agentic highlights reliability by routing queries to domain-specific endpoints instead of generic search results. One example used the ECB exchange-rate tool to convert 10,000 euros into dollars with a current reference rate, contrasting that with the risk of stale or hallucinated answers from ordinary lookup methods. The editorial takeaway is that agents backed by verifiable APIs can narrow one of generative AI’s biggest weaknesses: uncertain factual grounding. Pay-per-use access also hints at a cheaper alternative to stacking many niche software subscriptions.
AI productivity gap hits junior workers
A growing warning in digital work is not abstract displacement but a widening gap between automated operators and manual executors. Repetitive tasks such as posting content, editing simple videos, issuing invoices, scheduling campaigns and preparing quotes are described as especially exposed to workflow automation. A cited Stanford finding pointed to entry-level stress, with demand for workers aged 22 to 25 reportedly down 15% and then a further 19% afterward. The practical message is that routine software labor is being compressed fastest at the bottom of the experience ladder.
ChatGPT use no longer enough
The emerging divide is increasingly framed as automation versus assistance, not AI users versus non-users. Occasional queries to ChatGPT or Claude may speed research, but they do not remove repeated labor if workers still copy, paste and re-explain the same tasks every day. That makes superficial adoption risky, because people can overestimate how protected they are while competitors build actual workflows. In editorial terms, the winner is shifting from the best prompt writer to the best systems designer.
Kling 4.0 stretches Tapit ad
Kling 4.0 was used with Xfield to expand a concept for Tapit from a 10-second clip into a 30-second ad. The longer runtime allowed a more complete narrative arc around a creator turning audience attention into revenue, rather than a single visual gag. That makes the release notable less for raw image novelty than for practical ad-building flexibility. The experiment underscores how quickly generative video tools are moving toward usable branded storytelling formats.
Brand realism remains generative video's hurdle
The Tapit ad exercise also exposed familiar weaknesses in AI video production: continuity, realism and consistent brand control. Creative revisions pushed the opening toward a darker, more believable setup, including a creator with 20,000 followers but €0 in revenue before the turnaround, and reframed the lead as a European digital nomad. Those changes were meant to counter the glossy artificiality that often weakens generated ads. The result is a reminder that cheaper production is arriving faster than dependable cinematic coherence.
Videos covered
Grok Bot Full Tutorial for Beginners: How to Use & Set up Grok Bot
- •Cloud-based operation
- •Onboarding and workspace layout
- •Bots are defined as roles, not chats
Ton métier va changer dans 2 ans : prépare-toi maintenant (le setup simple)
- •Routine digital work is under pressure
- •Junior workers appear most exposed
- •The divide is no longer use versus non-use
Ce skill Claude Code met FIN à tes abonnements (payé à l'appel)
- •A tool layer instead of manual integrations
- •Fast setup with Claude Desktop
- •More reliable data than ordinary web search
On a testé le nouveau Kling 4.0 sur Higgsfield sur une vraie pub (le résultat ?)
- •From 10 seconds to 30 seconds
- •A revised storyline for Tapit
- •Push for a darker, more realistic opening