Latest news on AI development for June 17
Moonshot AI — a Chinese tech unicorn and creator of Kimi Chat — has released its new model Kimi-Dev-72B to the public. This 72-billion-parameter language model outperforms solutions like DeepSeek R1/V3 and Devstral in benchmark tests.
What does this mean?
• The team hosts the model on their own servers, implements updates, and tests them through automated systems. As a result, migrating old code takes hours instead of several days.
• New opportunities emerge for code analysis and security services: content generation has become free, but quality is still paid for separately.
OpenAI has integrated support for the Model Context Protocol (developed by Anthropic for data exchange between agents) directly into ChatGPT.
Why is this important?
Now developers can connect their CRM or business analytics to GPT with a single API call, without needing to create special plugins.
Example: a bank deploys an internal knowledge assistant without modifying production infrastructure.
Additionally, gateways with logging and audit functions will appear — a new subscription option for enterprise clients.
TikTok has expanded the capabilities of its advertising service Symphony AI: now available are features for converting images to videos, text to videos, and creating AI avatars directly within Ads Manager.
What does this mean?
• Marketers record a product with their smartphone, receive a ready-made video within a minute, and launch A/B testing.
• Business agencies shift from producing videos to optimizing prompts and analyzing conversions.
Reddit has introduced new tools — Insights and Conversation Summary — providing real-time analytics and automatic extraction of key discussion points for brands.
What does this change?
• Social media managers can respond more quickly to negative feedback without relying on third-party monitoring tools.
• Analytics startups will need to delve into detailed analysis or seek new platforms for their work.
Google plans to terminate its agreement with Scale AI after the startup received investments from Meta; meanwhile, Microsoft, xAI, and OpenAI are also seeking alternative partners.
What’s happening?
Re-tendering for data annotation contracts has been announced: this opens opportunities for new companies to enter the market and offer automated solutions.
For example: a small data-ops startup takes on a project processing 10 million images, while larger players reconsider their contractor chains.
DeepMind, in collaboration with creative studio Primordial Soup, showcased the short film “ANCESTRA” at the Tribeca Festival. The film combines live-action footage with scenes generated by the Veo model.
What’s interesting?
Producers save money on VFX production: Veo completes scenes without additional filming.
The development of AI-video licensing infrastructure (copyright management, scene marketplaces) is becoming an important part of the film industry.
Bonus: Geoffrey Hinton — one of the leading experts in AI — explained which professions will be the first to face automation. Primarily, these are fields with routine tasks: call centers, data entry, accounting, tech support — everything that relies solely on text or simple keyboard input. Physical labor — such as lifting, transporting, or manual work — will require robotics and sensors. Such jobs are likely to remain less affected by automation in the near future.
What can we infer from this?
It’s easier for businesses to upgrade their back-office systems using large language models than to hire new specialists. Now is the time to learn how to craft prompts and verify bot responses to stay competitive.
Investors should focus on services that replace manual text input: conversation transcription, basic contract drafting, or tech support. Long-term prospects are tied to robotics: robots capable of seeing and working with their hands require significant investment and are more complex to deploy but offer a chance to stay ahead of competitors.
In life as well: if your work involves only keyboard and copy-paste tasks — now is the perfect time to learn how to manage AI tools. If you’re involved in physical labor — keep an eye on developments in manipulators and robots to understand the first tasks they will be able to perform independently.