February 26, 2025
When I started building My Calendy, I dove deep into cognitive science research. What I discovered was fascinating: our mental abilities aren't static throughout the day. Instead, they follow predictable patterns that vary based on several factors. At My Calendy, we've synthesized research from various cognitive domains to identify eight key cognitive abilities that are particularly relevant to productivity and task performance – analytical, perceptual, creative, conceptual, strategic, administrative, technical, and collaborative. These represent practical categories that bridge the gap between scientific research and real-world application.
Research demonstrates that cognitive performance fluctuations aren't random but follow patterns influenced by our circadian rhythms, sleep patterns, and energy levels (Schmidt et al., 2007; Folkard & Tucker, 2003). Schmidt et al. (2007) specifically highlight how "circadian rhythms affect several cognitive processes" including attention, executive functions, and memory access. For example, research indicates that analytical problem-solving tends to be more effective during an individual's optimal time of day, which varies based on chronotype. As Wieth & Zacks (2011) found, participants performed better on analytical problems during their optimal time of day (morning for morning-types, evening for evening-types), whereas insight-based creative problem-solving actually showed advantages during non-optimal times. This pattern may occur because, as they explain, "at non-optimal times of day, decreased attentional control may allow a broader search of solution space, leading to increased solving of insight problems" (Wieth & Zacks, 2011, p. 399).
The core of My Calendy is the "Cognitive Timeline" – a personalized map of when your specific cognitive abilities peak and valley throughout your day. Building this wasn't simple. I started with baseline data from scientific studies showing general cognitive patterns across populations. But people aren't all the same.
The research clearly demonstrates that individual factors significantly impact cognitive performance patterns. Chronotype – whether someone is naturally more alert in the morning or evening – is particularly influential. Roenneberg et al. (2007) conducted extensive research on chronotypes, finding that they follow a normal distribution in the population and can significantly affect when individuals experience peak cognitive performance. Additionally, sleep quality and duration have profound effects on cognitive function, with Möller-Levet et al. (2013) demonstrating that insufficient sleep disrupts the expression of circadian rhythm-related genes, affecting biological processes that support cognitive function.
It's important to understand that different cognitive abilities respond distinctly to factors like sleep deprivation. Research by Lockley et al. (2004) found that reducing interns' work hours (thus increasing sleep time) significantly reduced attentional failures, with those working traditional 30-hour shifts making 36% more serious medical errors compared to those with reduced hours and more sleep. Interestingly, while certain analytical and attention-based tasks deteriorate with fatigue, Wieth & Zacks (2011) demonstrated that creative problem-solving can sometimes benefit from non-optimal times of day. Their study found that participants solved significantly more insight problems (requiring creative thinking) during their non-optimal time of day compared to their optimal time. My Calendy's system incorporates these nuanced relationships rather than applying simplistic adjustments across all cognitive domains.
One of our biggest technical challenges was handling all the scientific data efficiently. I used Supabase for our database needs, but I didn't just create a single database. I built several interconnected databases to handle different aspects of the cognitive timeline.
For storing and accessing the scientific research itself, I used vector databases. This allowed us to semantically search through research papers and extract relevant information. Then I created structured databases for the baseline cognitive patterns, chronotype adjustments, energy impact factors, and sleep duration effects.
The magic happens when these databases work together. When a user inputs their sleep duration, chronotype preference, and current energy levels, our system queries multiple databases simultaneously, applies the appropriate adjustments, and generates a personalized cognitive timeline for that specific user on that specific day.
While cognitive science tells us when we're best at certain types of thinking, project management science tells us how to break down projects into manageable tasks. I studied different project management methodologies—from Agile to Waterfall to Kanban—to understand how various projects are best structured.
When a user inputs a project, Cally first analyzes what type of project it is and which project management approach fits best. Then it breaks the project into phases, identifying which cognitive abilities are needed for each phase. For example, an initial brainstorming phase might require creative and conceptual abilities, while a review phase might need analytical and perceptual skills.
Once Cally knows both what needs to be done (project phases) and when you're best equipped to do specific types of thinking (cognitive timeline), she matches tasks to the optimal time slots while working around your existing calendar commitments.
A key design principle for us was to build Cally's cognitive intelligence model in-house. I wanted to create a system that leveraged our database of cognitive science and project management research while minimizing dependence on external large language models. This wasn't primarily about costs—it was about creating a specialized system tailored specifically to cognitive performance patterns and building a sustainable service that could scale.
Our approach was to develop our own cognitive intelligence algorithms and databases in-house, while using Langchain and OpenAI only for natural language understanding and assembling our analysis into coherent outputs. From the beginning, we hoped this approach would be more efficient than relying heavily on external LLMs, and our development process confirmed this.
When I first started building, we were sending too much information to OpenAI, with each request requiring 5,000-7,000 input tokens and 700-1,000 output tokens. But since OpenAI and Langchain are only piecing together the knowledge that Cally creates internally, we were able to refine our approach and dramatically reduce this to just 100-500 input tokens and 100-300 output tokens per request.
This 90% reduction in token usage was a welcome confirmation of our initial vision. By handling the cognitive analysis and scheduling logic in our own systems, we've built a service that's not only more affordable to maintain but also less dependent on external APIs for its core functionality, allowing us to offer subscriptions at just $5 per month.
I used Langchain to integrate these various components, creating a system where our database-driven cognitive analysis feeds precisely formatted context to the AI, resulting in focused, relevant responses without unnecessary token usage.
The current version of the Cognitive Timeline is just the beginning. Research reveals numerous factors influencing cognitive performance beyond our initial focus on chronotype, sleep, and energy levels. For example, West et al. (2002) demonstrated significant interactions between age and time of day, with older adults showing better performance on working memory tasks in the morning while younger adults showed less time-of-day variation. Similarly, Evans et al. (2017) found that typical university class schedules are misaligned with undergraduate students' peak cognitive times, suggesting that even institutional factors can impact when we perform best cognitively.
On the technical side, I'm working on real-time rescheduling capabilities that can quickly adapt when your day doesn't go as planned. This involves complex constraint solving algorithms that can efficiently reoptimize a schedule while considering dependencies between tasks and deadlines.
Looking forward, we're exploring the development of our own specialized language model based on the cognitive science and project management methodologies that form Cally's foundation. This research-driven approach would allow us to eliminate our reliance on external LLMs entirely, bringing all aspects of our service in-house. A dedicated model would understand the nuances of cognitive fluctuations and project planning at a deeper level, enabling more personalized recommendations without the overhead of explaining these concepts to a general-purpose model. The specialized nature of cognitive performance patterns and task management makes this an ideal domain for a focused LLM that embodies the specific research we've synthesized, rather than requiring us to extract relevant insights from broader models. This represents the natural evolution of our work to build a system that truly understands how to align human cognitive patterns with optimal productivity.
Building My Calendy has been a journey of combining cognitive science, project management principles, and modern web technologies. The result is a tool that doesn't just help you organize your time, but actually understands how your brain works throughout the day.
I've learned that the best productivity tools don't just tell you what to do—they help you do it at the right time. By understanding and working with your natural cognitive rhythms instead of against them, we can all achieve more without working harder. That's the science and technology behind My Calendy, and I'm excited to keep improving it based on new research and your feedback.