AI-Powered Workflow: A Founder's Approach to Maximizing Leverage

Y Combinator · 10.4K views

R-report

• Circleback, an AI note-taker, acts as a "company brain" by capturing, organizing, and automating tasks from conversations. • The founder uses a minimalist hardware setup, including a specialized keyboard and AirPods Max to signal focus. • Beyond its core function, Circleback is leveraged as an Applicant Tracking System (ATS) and CRM, and for personal interview prep. • The engineering workflow is shifting from manual coding to orchestrating AI agents and reviewing their output. • The future of software engineering emphasizes imagination and strategic building over raw coding speed, with comprehensive recording becoming essential.

🧠 The AI-Powered "Company Brain"

Circleback, an AI note-taker, is designed to capture and organize every aspect of a company's conversations. It records and transcribes meetings, automatically generates notes, and assigns action items. This centralized knowledge base allows users to build automations that pull customer information, update CRMs, issue trackers, and even Slack channels. Essentially, Circleback aims to be a "company brain" that remembers everything, enabling users to query past conversations and connect various business applications.

💻 A Minimalist Yet Optimized Setup

The founder maintains a minimalist desk setup, prioritizing functionality and personal preference. His keyboard is a Topre (Tokyo Precision) membrane keyboard, chosen for its satisfying typing feel and quiet operation, which is ideal for an open office environment. A small blue toy car, acquired through a "hypnotizing" Temu onboarding process, serves as a fidget toy during discussions. For audio, he uses AirPods Max, which also double as a non-verbal cue to the team that he is "locked in" and focused on deep work.

🚀 Beyond Meetings: Unexpected AI Applications

Circleback's capabilities extend far beyond standard meeting management:

• **Applicant Tracking System (ATS):** The founder uses Circleback to manage the recruiting pipeline. He has a dedicated view for "Interviews" in the action items, allowing him to track follow-ups with candidates. Each candidate has a "person page" that aggregates all touchpoints, including emails and meetings, providing a comprehensive timeline of their interactions with the company.

• **CRM and Company Knowledge Base:** Circleback functions as an internal CRM, tracking interactions with people and companies. For instance, the Y Combinator company page displays all contacts, outstanding action items, and a full timeline of conversations, aiding in strategic engagement and follow-up.

• **Personal Interview Preparation:** The founder leverages Circleback to prepare for interviews. It analyzes past emails and conversations related to the interview topic, pulls logistics from his calendar, and even suggests cool product features to showcase based on the anticipated questions.

• **Mobile and Wearable Integration:** With the iPhone's Action Button configured to start a Circleback recording, quick voice notes can be captured effortlessly. A forthcoming Apple Watch app will further streamline this process, allowing recordings to be initiated with a single tap on a watch complication, ideal for capturing thoughts while on the go.

🛠️ Engineering Workflow in the AI Era

The founder's daily workflow reflects a significant shift towards AI-driven development. While he still engages with coding tools, a growing portion of his time is dedicated to reviewing code and orchestrating AI agents. The goal is for Circleback's internal tools to be "always on" and autonomously handling tasks. Specialized AI agents manage different company functions, from finance to customer support, centralizing these processes in Slack channels for team visibility and collaboration. This approach allows engineers to focus on high-leverage activities, with dedicated "no meeting days" for deep work and end-to-end feature shipping.

✍️ Crafting AI: Prompt Engineering and Quality Control

Developing effective AI agents involves rigorous prompt engineering and quality control. The team employs "evals" for quantifiable tasks, ensuring accuracy and consistency. A key focus is on making AI-generated notes sound human-written and avoiding redundant phrases (e.g., never using "discussed" in meeting notes). This iterative process involves updating models and prompts, constantly measuring performance to ensure incremental improvements without introducing new issues. The philosophy is to automate tasks that "no one should be doing," freeing up human talent for more complex and creative work.

🚀 The Future of Software: Imagination Over Bandwidth

The founder believes that the current era is an incredibly exciting time for software development. As Large Language Models (LLMs) and AI agents become more capable, the traditional constraints of skill and bandwidth are being replaced by the limits of imagination. The cost of writing software is decreasing, shifting the focus from "how fast you ship" to "what you decide to build and how you decide to build it." The ability to record and share the full context of all company conversations is becoming paramount, as it empowers AI agents and human teams alike to make informed decisions. This environment, especially within a startup, fosters rapid experimentation with cutting-edge tools, unburdened by legacy systems and bureaucratic "red tape."