Welcome To The Age of Deep Research
How OpenAI, Perplexity, and Google are unlocking human productivity
Remember the last time you spent hours diving into rabbit holes of browser tabs, desperately trying to piece together information for an important project?
Those days are behind us.
A new breed of AI tools is emerging that will transform how we conduct research and process information.
Leading tech companies like OpenAI, Perplexity AI, and Google are ushering in the age of "deep research AI assistants" – sophisticated tools that don't just search, but understand, synthesize, and explain complex information in ways that were previously impossible.
From Search To Deep Research
Unlike traditional search engines that simply match keywords and return links, these AI assistants act more like skilled research partners. They can understand nuanced questions, synthesize information from multiple sources, run comprehensive searches and present coherent, cited answers that save hours of manual research time.
Take Perplexity AI, my daily driver. Last week they released “Deep Research” capabilities as part of their Pro subscription. Deep research leverages the reasoning capabilities of LLMs (Large Language Models) to iteratively refine and research answers to user questions.
Here is an example of Perplexity Pro’s Deep Research in action:
As you can see, Perplexity AI LLMs will reason through information step by step until they arrive at a comprehensive answer to the user question.
These AI companies all offer a form of deep research today:
OpenAI Deep Research (USD 200/mo for Pro users, limited to 100 queries/mo)
Perplexity Pro Deep Research (USD 20/mo, requires a monthly subscription)
Google Deep Research (USD 20/mo for Gemini Advanced users)
As you can see, Perplexity isn’t the only vendor offering this capability, but because they are using open source models like DeepSeek’s R1 they can offer the same agentic AI capabilities that cost USD 200/mo at OpenAI for USD 20/mo — quite the price differential!
From Months To Weeks To Days To Minutes
The impact on my personal productivity has been staggering.
I’ve been working with Perplexity’s Deep Research for the past week and it is the AI product of the year for me so far — it’s on par with how Cursor IDE changed my coding practice completely back in August 2024.
You can ask Deep Research assistants to produce undergraduate-level essays and syntheses in a matter of minutes.
If I had to do this manually in a library this would take weeks or months!
And even with the help of the most advanced search engines, gathering a complete picture on a subject would easily cost days of intense bio-cognitive research before.
So it’s no surprise that I think these kind of AI tool will have a massive impact on research and academia.
Especially since a lot of research chairs only exist because of the amount of time and bio-cognitive effort needed to gather a half-decent understanding of a scientific field.
Google actually made a big step in this direction this week when Google Deepmind announced their “AI co-scientist” — an AI system designed to help drive speed up scientific discoveries:
A Different Kind Of Bottleneck
It’s not all sunshine and roses though — these AI tools rely on high-quality sources.
And as anyone who’s surfed the internet will be aware, not all data is created equal.
So even though the information-processing capacity of the human brain are quickly become less of a bottleneck, before these tools can become real agents of change a lot of work will needed to unlock and integrate the right data sources.
As an aside — if you’re interested in creating Deep Research capabilities for your organisation, the team at Lodestone Digital can help. Check out our services or reach out if you want to know more!
The Bottom Line
The age of deep research AI assistants is here, and it's transforming how we work with information. While these tools won't replace human expertise, they're becoming indispensable partners in our quest for knowledge and understanding.
Your First Robot Could Speak Chinese: How Unitree's Humanoids Are Leading The Charge
Forget about Tesla Optimus and Boston Dynamics for a minute.
Imagine walking through a factory where tireless robotic assistants glide between assembly lines, their sensors scanning equipment with laser precision while human supervisors focus on strategic decisions.
This isn't science fiction—it's happening today in Chinese manufacturing hubs through Unitree Robotics' groundbreaking H1 humanoids.
The Rise of the Machines (That Actually Work)
Unitree has been quietly revolutionizing the humanoid robot space with a laser focus on what matters most to businesses: practicality and affordability.
Their H1 model represents a paradigm shift—a 5'6" robotic workforce member costing less than a luxury sedan. And this machine solves concrete problems:
Precision Without Fatigue: Operating 22-hour shifts inspecting circuit boards at Foxconn facilities, achieving 99.97% defect detection accuracy.
Danger Defier: Scaling smoke-filled refinery structures in Shanghai, performing thermal inspections that previously caused 12% annual worker injury rates.
Logistics Game-Changer: Lifting ~40kg engine blocks at BYD automotive plants with millimeter positioning repeatability.
"What makes Unitree revolutionary isn't the technology—it's the economics. … They've achieved what Honda's ASIMO couldn't—a humanoid form factor at industrial robot prices."
—Dr. Li Wei, Zhejiang Robotics Institute.
Tomorrow's Workforce, Deploying Today
While analysts debate humanoid viability, early adopters report tangible impacts:
Jinjiang Manufacturing Consortium saw
34% reduction in quality control staffing costs
19% increase in nightly production throughput
ROI achieved in 8.2 months through accident prevention
These aren't hypothetical projections—they're boardroom metrics from factories currently scaling Unitree deployments.
The secret lies in targeted augmentation rather than outright replacement. At Shenzhen's Haier Smart Factory, human supervisors now manage six robotic teams simultaneously, focusing on process optimization rather than repetitive checks.
The Adoption Timeline Accelerates
And Unitree isn’t messing about.
Within two years, jobs that require physical labor will no longer look the same:
12-18 Months Out:
Pilot programs available through Siemens' automation partners
Customizable API integration for legacy manufacturing systems
Subscription models starting at $3,200/month per unit
2027-2028 Horizon:
Industry-specific variants (pharmaceuticals, semiconductor, construction)
Swarm intelligence enabling collaborative robot teams
The Smart Money Move
Here's where it gets interesting for business leaders:
Early Mover Advantage: If you're in manufacturing, logistics, or quality control, start exploring pilot programs now. Being first in your industry to successfully implement humanoid robots could give you a serious competitive edge.
Workforce Evolution: This isn't about replacing workers – it's about augmentation. Start planning how your team could leverage these robots to enhance productivity. Think: What could your people achieve if freed from repetitive tasks?
Supply Chain Innovation: Imagine reshoring operations becoming economically viable again, or your warehouse operating 24/7 with minimal downtime. The possibilities are game-changing.
The Bottom Line
While the exact ROI figures are still emerging (it's early days, after all), the potential for transforming operations is clear. The key is starting small, focusing on specific use cases where traditional automation falls short, and building from there.
This week in AI
Google DeepMind announced their AI Co-Scientist: Co-scientist is an AI system designed to accelerate scientific discovery. It can formulate hypotheses, design experiments, analyze results, and even suggest new avenues for exploration, potentially dramatically speeding up research across various scientific disciplines.
Microsoft’s Majorana chip flips the bits: the architecture of Microsoft’s new quantum computing chip announced earlier this week brings us one step closer to fault-tolerant quantum computers. Their work focuses on creating stable qubits, the fundamental building blocks of quantum computers, which are crucial for overcoming errors that have plagued the field.
xAI released Grok 3 in beta. Their latest LLM boasts advanced reasoning capabilities and performs well on tasks like math, coding and multimodal understanding. In addition, their integration with X means they have access to live text updates of events and opinions worldwide making it an interesting addition to the AI assistant playing field.