OpenAI just shared an internal five-level framework to track its progress toward AGI, and executives say it is already closing in on Level 2. The ladder runs from basic conversational chatbots up to "Reasoners" capable of human doctorate-level problem-solving, all the way to systems that can do the work of an entire organization. It gives the clearest view yet of how the company defines human intelligence.
Google DeepMind just hooked up Gemini 1.5 Pro to its discontinued Everyday Robots hardware, turning them into autonomous office navigators. The bot can process spoken commands and even follow hand-drawn whiteboard maps across a 9,000-square-foot office, notching a 90% success rate across 50 employee interactions.
Shanghai AI Lab just open-sourced InternLM 2.5-7B, cramming a massive 1M token context window into a lightweight architecture. The model claims top scores for sub-12B models on Hugging Face's Open LLM Leaderboard, bringing heavy-duty reasoning and tool use to local hardware.
Perplexity just upgraded Pro Search to handle complex questions using multi-step reasoning. The search assistant can now break down queries, trigger follow-up web searches automatically, and run code or tap Wolfram|Alpha for advanced math and data analysis. Free accounts get limited access, while Pro subscribers get higher daily limits.
LangChain just rolled out the stable release of LangGraph v0.1 alongside a closed beta of LangGraph Cloud. The framework ditches rigid abstractions to give developers low-level control over multi-agent workflows, complete with human-in-the-loop approvals, time-travel debugging, and dedicated infrastructure for scaling agent fleets reliably.
Google just opened up Gemini 1.5 Pro's 2-million-token context window to all developers, removing the waitlist. Both 1.5 Pro and Flash now also feature built-in Python code execution in an offline sandbox to tackle math and data reasoning, alongside Gemma 2 landing in Google AI Studio. Why it matters: developers can now feed entire codebases into Gemini while letting the model verify its own math by running code.
Reverse-engineering group Rabbitude just revealed that Rabbit left critical API keys hardcoded in its codebase, allegedly allowing anyone to access every response the R1 has ever given. The keys could also let attackers brick devices or alter responses. Rabbit stated it was alerted on June 25 and has found no evidence of compromised user data so far.
Zeta Labs, founded by former Meta engineers, just launched Jace, an autonomous AI agent capable of executing complex multi-step tasks inside a web browser. Powered by their proprietary AWA-1 model, Jace handles everything from booking hotels to setting up ad campaigns on command, backed by a fresh $2.9M pre-seed round.
Sussex businessman Steve Endacott just launched "AI Steve" to run in next month's UK general election. Powered by Neural Voice, the AI avatar chats with voters 24/7 via voice and text, while Endacott acts as its physical proxy in Parliament voting strictly based on constituent feedback. Why it matters: It is a radical experiment in continuous direct democracy, testing whether AI can actually make politicians more responsive.
Anthropic just made tool use generally available across the entire Claude 3 model family via its API, Amazon Bedrock, and Google Cloud's Vertex AI. Developers can now connect Claude to external APIs to extract structured data, automate multi-step workflows, and stream tool interactions. Why it matters: It gives Claude the hands and feet needed to act as a true agent, though parallel tool calls aren't supported yet.
OpenAI just opened custom GPTs, data analytics, and vision features to all free ChatGPT users. Anyone can now explore and chat with community-made bots or analyze files from cloud storage, though building your own GPT still requires a paid tier. Once free users hit message caps on GPT-4o, conversations automatically drop back to GPT-3.5.
Elon Musk just doubled down on his aggressive AI timeline, replying on X that AGI will arrive "next year." While Musk continues to push for rapid breakthroughs at xAI, experts remain split on whether true AGI is anywhere close to ready.
Near is developing an AI agent designed to help non-coders build entire decentralized applications (DApps) simply by describing what they want. Co-founder Illia Polosukhin said the upcoming tool will generate smart contracts, frontends, and middleware end-to-end, pitching user-owned open-source AI as a counterweight to centralized tech giants.
Microsoft just introduced Team Copilot at its Build conference, expanding its generative AI tech from personal assistance to full team collaboration. Embedded across Teams, Loop, and Planner, it can take shared meeting notes, summarize chats, assign tasks, and track deadlines. It launches in preview later this year for Copilot for Microsoft 365 subscribers.
OpenAI just opened its custom GPT Store to all free ChatGPT users, unlocking access to millions of tailored bots previously kept behind a paywall. The move massively widens the audience for GPT creators, though you'll still need a paid subscription if you want to build and publish your own.
Bumble founder Whitney Wolfe Herd just shared a wild vision for the future of online dating at the Bloomberg Technology Summit: personal AI concierges that interact with hundreds of other bots to find your top matches. Instead of making small talk with 600 people, your agent could scan the city and narrow it down to three candidates you should actually meet in real life.
Hugging Face just launched LeRobot, an open-source robotics library designed to act as the "Transformers for robotics." The toolkit includes pretrained models, simulation integrations, and lightweight dataset tools to control hardware from robotic arms to humanoids. Why it matters: Hugging Face is betting on open-source and crowdsourced data to prevent big tech from monopolizing physical AI.
GitHub just launched the technical preview of Copilot Workspace, a Copilot-native developer environment designed to build software from scratch in natural language. Starting from an issue or repo, agents brainstorm, plan, and test implementation while leaving developers in full control of every step. Why it matters: Coding is shifting rapidly from managing syntax to expressing intent.
Meta advertisers are watching budgets rapidly evaporate as the company's automated platform, Advantage Plus, misfires with skyrocketing costs and dismal returns. Multiple marketers reported CPMs spiking up to 10 times higher than normal while generating virtually zero revenue. Why it matters: The promise of 'set it and forget it' automation is colliding with costly bugs, driving frustrated businesses back to manual ad setups.
OpenAI just rolled out enterprise-grade controls for its API customers, introducing a Batch API that cuts prices by 50% for non-urgent tasks completed within 24 hours. The Assistants API also received a massive boost, expanding its file search limit by 500x up to 10,000 files alongside native streaming and project-level workspace controls.
