Notes from the workshop

Arga Labs has secured a $10 million seed round led by General Catalyst, with backing from Box Group, Emergence, Gradient and SV Angel. The funding signals continued investor interest in tools for building and training enterprise AI agents.

QueryStory has launched out of stealth with $6 million in seed funding and a plan to combine large language models with cybersecurity expertise. The goal is to make AI-generated queries more coherent and easier to trust for developers and Telegram users alike.

Bluesky has expanded video support to let users upload clips up to 10 minutes long, alongside faster upload speeds. The update matters for creators and platform builders watching how social apps evolve their media features.

Robot bodies are advancing, but their AI brains are still catching up. A TechCrunch report looks at how robot developers are moving beyond earlier-generation models to improve real-world performance.

Meta has settled an $18 billion lawsuit brought by 29 states over claims that its platforms were designed to be addictive to children despite known risks. The case highlights growing legal and regulatory scrutiny of social media products and their impact on young users.

Z.ai has confirmed it is the lab behind Ox Alpha, the open AI model that has been climbing benchmarks and leaderboards. The company says the model weights will be released soon, adding a new development to watch for Telegram and software developers.

The U.S. cyber agency says hackers targeted over 100 water systems during July, as suspected Iran-backed activity continues to hit critical infrastructure. The incident underscores how exposed essential services remain to coordinated cyber operations.

Bill Gates is still broadly aligned with responsible AI efforts, but TechCrunch says he is now backing a few newer ideas, including a robot tax and jobs reserved for humans. The proposal highlights how the debate over AI’s labor impact is moving beyond efficiency into policy and workforce protections.

Ring has introduced a new encryption standard and is now using it by default for cloud features, while still allowing users to opt in to end-to-end encryption. The change is notable for anyone tracking consumer security and the tradeoffs between convenience, privacy, and cloud functionality.

Perceptron is pitching an AI model designed to help machines navigate the world while also delivering deeper visual intelligence. The effort points to growing interest in using visual AI in industrial settings.

A lot of Telegram automation fails not because the bot logic is weak, but because developers treat Telegram like a generic messaging API. The most common mistake is designing automation around a single happy path instead of Telegram’s real event model, permission rules, and delivery quirks.

Telegram gives developers a fast, high-signal channel for AI experiences, but the right implementation depends on whether you need conversational automation, retrieval, moderation, or user-facing agents. This comparison breaks down the most useful integration patterns, tradeoffs, and product opportunities for Telegram developers.