Perplexity’s New Labs Feature Just Replaced Your Research Team
This is the best AI consultant I've ever used.
Good morning AI entrepreneurs & enthusiasts,
Today’s drop is loaded. Perplexity just leveled up from search bar to straight-up consulting killer—it’s cranking out dashboards and go-to-market strategy decks that would cost six figures from McKinsey. Meanwhile, DeepSeek R1 didn’t inch forward—it rocket-jumped into the top 3. R2’s not just coming... it’s looming, and it's looking like it could eat the leaderboard alive.
Zochi went full autodidact and dropped a peer-reviewed paper with zero human co-authors—accepted at ACL 2025, no less. And the UK + EU? They’re squashing the beef with a shared supercomputing pact to make Europe a serious contender in the global AI race.
In today’s AI news:
DeepSeek R1 jumps to #3 on the LLM leaderboard
Perplexity Labs launches Pro-only multi-tool workflows
Zochi becomes first AI to publish peer-reviewed science
UK & EU partner on supercomputing-powered AI innovation
Top Tools & Quick News
Perplexity Labs launches multi-tool workflows for Pro users
The News: Perplexity Labs is a new feature exclusive to Perplexity Pro subscribers, offering AI-powered automation to streamline research, reporting, and project creation.
The details:
Multi-tool automation: Combines web search, data analysis, code execution, and chart generation into one seamless environment.
Project-ready output: Build reports, spreadsheets, dashboards, mini apps, and more, all in response to a single prompt.
Longer, deeper sessions: Labs invests 10+ minutes per project for more detailed and refined output.
Asset organization: Charts, documents, and code are automatically saved to an “Assets” tab.
One-stop creation environment: Handle the full workflow from research to delivery in one platform.
Why it matters: Labs is more than a productivity tool—it's a glimpse into the future of AI-assisted work. By automating multi-step research and creation workflows, it empowers users to go from concept to deliverable without juggling tools. From financial models to marketing plans, Labs reduces the overhead of high-output work. Its synthesis of data, visualizations, and finished deliverables rivals what you'd expect from top-tier consultancies like Deloitte and McKinsey—only faster, cheaper, and infinitely scalable.
DeepSeek R1 leaps to #3 on LLM leaderboard
The News: DeepSeek R1 has surged to the #3 position on the Artificial Analysis leaderboard, following its latest R1-0528 update in May 2025, marking a new milestone in the open-source LLM race.
The details:
On top benchmarks like GPQA, ADER, and LiveCodeBench, DeepSeek now matches or outperforms xAI’s Grok 3 mini and Meta’s Llama 3.
DeepSeek R1 is the top open-weight model globally—MIT licensed with full commercial rights.
Key specs: 685B parameters (37B active), 128k token context, $0.96 per 1M tokens blended, 31.9 tokens/sec output, 3.15s TTFT.
Why it matters: This wasn’t a minor update—it was a leap. DeepSeek’s massive 8-point gain on the leaderboard rivals some of the biggest step-changes we’ve seen in LLM development. The speed, accuracy, and cost efficiency of R1 have made it the definitive leader in open-weight models. But more importantly, this explosive growth raises a critical question: If R1 just made this kind of jump... what’s coming with R2? The landscape may be about to shift again—faster than anyone expected.
AI learns first peer-reviewed publication
The News: Zochi, developed by Intology, has become the first AI system to independently publish a paper accepted at the main proceedings of ACL 2025, one of the most selective conferences in natural language processing.
The details:
Zochi completed every stage of the research process autonomously—from literature review and hypothesis generation to experimentation, analysis, and manuscript writing—requiring human assistance only for minor formatting.
The paper was accepted into ACL's main track, scoring in the top 8.2% of all submissions with a 4.0 meta-review rating.
This milestone surpasses previous achievements like Sakana AI’s AI Scientist-v2, which only cleared workshop review standards at ICLR.
Zochi will be released in beta to support collaborative research in autonomous hypothesis development.
Why it matters: Zochi's breakthrough marks a turning point in scientific research. For the first time, an artificial system has independently created and published a peer-reviewed paper at the level of top human researchers. As Intology’s announcement describes, this could fundamentally accelerate discovery, reduce research costs, and broaden participation in high-impact science.
UK-EU announce landmark AI collaboration
The News: The UK and EU have formalized a new AI partnership aimed at accelerating joint research and development efforts by connecting UK public research institutions with Europe’s leading supercomputing infrastructure.
The details:
UK institutions can apply to host an “AI Factory Antenna” to access European compute power through EuroHPC Joint Undertaking, a consortium of 35 member countries.
Up to €5M in joint funding will support the UK-based facility, with matched investment from both sides.
The initiative will support high-priority research in healthcare, energy, and climate innovation.
Why it matters: This partnership represents a major post-Brexit collaboration with the potential to significantly advance AI breakthroughs by providing UK researchers and startups access to Europe’s highest-performance compute infrastructure. Beyond science, it’s a strategic move toward realigning regulatory standards and tech diplomacy, reinforcing the UK’s role in the global AI race.
Today's Top Tools:
DeepSeek R1 – Boosted reasoning performance
Neon – Opera's new AI-first browser
Retool Agents – Turn APIs into usable tools for AI agents
Quick News:
Tencent released HunyuanVideo-Avatar, a multimodal model that animates still images with audio and emotion control. It supports multiple characters and is fully open-source.
Hume's EVI 3 speech model beat GPT-4o in blind tests on emotion, expressiveness, and naturalness. It offers voice-to-voice interaction with rapid, multilingual support.
Resemble AI launched Chatterbox, a voice cloning model that outperformed ElevenLabs in tests. It features open-source deployment, real-time zero-shot cloning, and emotion exaggeration.
Manus debuted Slides, a tool that turns a single prompt into polished slide decks. It supports PDF input, real-time editing, and export-friendly presentations.
Thanks for reading this far! Stay ahead of the curve with my daily AI newsletter—bringing you the latest in AI news, innovation, and leadership every single day, 365 days a year. See you tomorrow for more!