Talk Python To Me
By Michael Kennedy
Talk Python to Me is a weekly podcast hosted by developer and entrepreneur Michael Kennedy. We dive deep into the popular packages and software developers, data scientists, and incredible hobbyists doing amazing things with Python. If you're new to Python, you'll quickly learn the ins and outs of the community by hearing from the leaders. And if you've been Pythoning for years, you'll learn about your favorite packages and the hot new ones coming out of open source.
35 episodes
All Episodes
Integrating Rust into Python via Stack-Native Veneer Architectures
Use Rust within Python to build high-performance, memory-safe systems while keeping your development process agile. By moving compute-heavy tasks to a compiled core, you prevent production crashes and keep the flexibility of your current codebase.
View Episode Notes →
Reducing Metadata Latency Through Transactional Catalog Architectures
Modern data lakes often face a performance penalty because of too many metadata round trips. If you treat your catalog as a transactional database instead of a simple collection of files, you can increase throughput by 40 times.
View Episode Notes →
Overcoming Python's Multiprocessing Tax With Multi-threaded Runtimes
Python's single-threaded asynchronous model imposes a hidden multiprocessing tax that consumes excessive memory and fails to utilize modern multi-core hardware. Learn how to transition to free-threaded architectures to reclaim lost CPU capacity.
View Episode Notes →
Replacing Manual Excel Workflows With Automated Research Infrastructure
Relying on Excel creates a bottleneck that slows down medical research. By replacing manual spreadsheets with agentic AI and structured infrastructure, research teams can automate discovery and gain a significant competitive advantage.
View Episode Notes →
Building Systemic Resilience Through Automated Infrastructure and Inoculation
Easy cloud choices often lead to hidden technical debt, turning small bugs into middle of the night outages. Build systemic resilience by replacing manual, custom server management with automated, disposable infrastructure and intentional failure simulations.
View Episode Notes →
Leveraging Django Async for Surgical Performance and Operational Efficiency
Django's asynchronous support is not a dangerous feature gap. Instead, it is a mature tool for achieving surgical performance gains. Stop over-provisioning your servers and start using async sidecars to boost efficiency without rewriting your entire application.
View Episode Notes →
Moving AI Data Agents From Static Files To Live Kernels
Treating notebooks as static files forces AI agents to guess your data state, which leads to frequent errors. Connect your agent directly to the live Python kernel to eliminate manual bookkeeping and enable truly reproducible data analysis.
View Episode Notes →
Building Systemic Guardrails for Reliable High-Stakes AI
Reliability in high-stakes AI comes from building systemic guardrails instead of just chasing model intelligence. Focus on the unglamorous infrastructure of grounding and refusal logic to stop probabilistic errors from turning into catastrophic failures in your production systems.
View Episode Notes →
Building Durable Competitive Advantage Through Custom Developer Tooling
Generic AI tools create a developer moat that limits your agency. You can gain a durable competitive advantage by building opinionated, terminal-first ecosystems that prioritize local control, composable toolchains, and adversarial verification.
View Episode Notes →
Managing Complexity in the Era of AI-Generated Code
AI-driven coding has made software cheaper to build, which means the ability to curate and verify code is now more valuable than writing it. Master the art of architectural oversight to navigate the flood of automated noise.
View Episode Notes →
Building Infrastructure Layers to Support AI-Native Application Architectures
Successful AI startups are moving past simple model wrappers. They are building infrastructure layers that enforce determinism and security. Learn how to architect agent-friendly systems that prioritize developer experience instead of manual orchestration.
View Episode Notes →
AI Amplifies Open Source Challenges: Maintainers Face Overwhelming, Low-Quality Contributions
AI amplifies open-source challenges, overwhelming maintainers with unmanageable code. Understand these shifts to safeguard project sustainability and shape the future of development.
View Episode Notes →
Dual-Audience Documentation: Bridging Human Readability and AI Accessibility
Documentation must now serve both humans and AI. Future-proof your projects by structuring content for machine readability, ensuring discoverability and integration in an AI-driven landscape.
View Episode Notes →
Event Sourcing: Unlocking Insights and Durable Systems with AI
Capture every system change as an immutable event, transforming your database into a powerful audit log for deeper insights and durable, AI-ready software.
View Episode Notes →
Ray: Orchestrating Scalable Python AI Workloads
Unlock scalable AI by transforming complex workloads into manageable Python tasks. Ray orchestrates advanced AI, from GPT-3 training to reinforcement learning, simplifying distributed systems for developers.
View Episode Notes →
Reclaiming Operational Ownership Through Strategic Self-Hosting
Self-hosting turns your digital infrastructure from a rented service into an asset you own. By replacing convenient cloud dependencies with personal administrative control, you build a system that is resilient, private, and permanent. This approach ensures your setup survives corporate policy changes.
View Episode Notes →
2025 OWASP Top 10: Shifting Security Beyond Code to Ecosystem
Modern software security demands a holistic view, addressing systemic vulnerabilities and design flaws alongside code. Understand the expanded attack surface and build resilient systems from the ground up.
View Episode Notes →
Python Packaging's "Fast" Approach Creates Future Performance Bottlenecks
Current Python packaging sacrifices future performance for immediate compatibility, creating bloated binaries and slower execution. Unlock significant performance gains by embracing specialized, hardware-aware package distribution.
View Episode Notes →
Deep Agents Harness LLMs for Complex Task Planning and Execution
Deep agents move beyond prompts, offering planning, file access, and sub-agent delegation for complex, long-running tasks. Understand their architecture to build truly capable AI assistants.
View Episode Notes →
Zensical: Modular, Performant Static Site Generation Beyond MkDocs
Rebuild from scratch to escape limitations: Zensical offers a modular, performance-driven platform, not just an incremental fix, for future-proof systems.
View Episode Notes →
Monty: Deliberately Limited Python for Secure AI Agent Code Execution
Monty offers near-instantaneous, secure Python execution for AI agents by embracing deliberate limitations, unlocking unprecedented automation and efficiency.
View Episode Notes →
Apache Airflow Monorepo: Modern Tooling Unlocks Development Advantages
Monorepos now offer unprecedented code isolation and developer efficiency, dissolving trade-offs between monolithic and multi-repo structures through advanced packaging and tools like `uv`.
View Episode Notes →
Python Typing Council: Shaping Semantic Interpretation of Code
The Python Typing Council is quietly defining how type checkers interpret code, evolving Python's type system dynamically through community proposals and practical needs.
View Episode Notes →
Digital Humanities Applications Face End-of-Life Problem After Funding
Digital archives face an "end-of-life" problem when funding ends. Transform dynamic web applications into static sites to ensure long-term access and eliminate prohibitive maintenance costs.
View Episode Notes →
DataStar: Server-Driven Web Development Simplifies Applications
DataStar's 11KB framework challenges web development norms, enabling real-time, high-performance apps with minimal JavaScript by prioritizing server-driven updates and native browser capabilities.
View Episode Notes →
FastAPI Cloud Manages Complexity for Pythonic Developer Velocity
FastAPI Cloud unlocks developer potential by expertly managing cloud deployment complexity, allowing you to focus on essential application logic and gain a significant competitive advantage.
View Episode Notes →
PyView: Python-Centric Real-Time Web Applications Via LiveView Protocol
Build real-time, interactive Python web apps without complex JavaScript. PyView leverages server-side state management and efficient updates for dynamic UIs, simplifying development.
View Episode Notes →
DiskCache: Persistent, Cross-Process Caching for Python Applications
Accelerate Python apps by persisting expensive computations to disk. DiskCache rivals network caches like Redis for local speed, reducing redundant operations and cloud costs.
View Episode Notes →
Python Web Deployment: Performance, Concurrency, and Framework Trade-offs
Upgrade Python versions for nearly double performance; leverage background tasks and optimize database indexing to boost responsiveness and efficiency.
View Episode Notes →
Python's 2025: Concurrency, Tooling, AI, and Funding Challenges
Python's future accelerates with the impending GIL removal and advanced tooling, promising 10x performance gains and simplified development, while the community navigates AI's impact and funding challenges.
View Episode Notes →
Right-Sized Python Deployment: Autonomy, Cost-Effectiveness, and Simpler Infrastructure
Deploy Python applications with right-sized engineering to cut cloud costs and complexity. Gain autonomy and simpler infrastructure management by running on a single server with Docker isolation.
View Episode Notes →
Standardizing Interactive Python Widgets With "Just Enough JavaScript"
Democratize interactive Python notebooks with "just enough JavaScript" for seamless, bidirectional data exploration across any environment.
View Episode Notes →
Demystifying Computing Through Building Interpreters and Emulators
Demystify programming by building interpreters and emulators. Discover how minimal syntax solves complex problems and understand the software-hardware interface, fostering adaptable problem-solving skills.
View Episode Notes →
Treating LLMs as APIs Requires Defensive Programming and Structured Output
Treat LLMs as programmable APIs, not black boxes. Implement defensive programming with caching and structured outputs to manage unpredictability and ensure data integrity.
View Episode Notes →
Model Context Protocol: Universal Interface for Agentic AI Integration
Connect any AI to any application or data with a universal interface, enabling dynamic, real-world interactions and composable agentic workflows beyond static training.
View Episode Notes →