17 Nov 2026 You cant blame AI for the code you ship | Freshworks Murali Swaminathan
Freshworks uses an AI Product Development Lifecycle to manage code quality and prioritize measurable customer impact over mere tool adoption.
Published 20 Mar 2026
Duration: 1912
Strategies for optimizing AI token usage, critiques of MCPs, context anchoring for memory degradation, debates on ephemeral vs. cumulative workflows, decentralized web practices, AI content misattribution, Apex framework for productivity impact, and a potential digital Renaissance via low-cost tools and AI integration.
Are rolling token blackouts and late-night AI coding shifts about to become the new normal for developers? This week on the Friday Deploy, Andrew and...
The podcast discusses strategies for optimizing token usage in AI systems, such as shifting workloads to off-peak hours (nights/weekends) to reduce costs and system strain, and exploring speculative ideas like "token blackouts" or storing tokens for later use. It critiques the overhyped rise and fall of MCPs (Microservice Component Packaging) as a tool for consistent AI agent workflows, noting their decline in favor of simpler CLIs (command-line interfaces) and the limitations of MCPs as a one-size-fits-all solution. The episode also delves into AI-assisted development challenges, emphasizing the need for context anchoringexternalizing decisions and reasoning into a "living feature document" to combat memory degradation in long coding sessions. Tools like Opus 4.6 (with a 1 million token context window) and workflows that balance ephemeral (short-lived, task-focused) and cumulative (archived, knowledge-heavy) approaches are highlighted as strategies to enhance AI collaboration and reliability.
The discussion extends to broader themes, including the resurgence of the "Small Web"a return to decentralized, user-controlled websitesas a counter to modern internet monopolies, and reflections on early web culture. It addresses pitfalls in AI usage, such as the "sloppy pasta" phenomenon, where unverified AI-generated content is shared without accountability, eroding trust. Best practices for AI integration include trimming verbose outputs, validating factual accuracy, and aligning AI workflows with human expertise. The podcast introduces the Apex Framework as a model for operationalizing AI in software engineering, focusing on measurable delivery outcomes (e.g., pull request success) rather than vanity metrics. It underscores the importance of systemic analysis of software development bottlenecks, ensuring AI tools complementrather than exacerbateexisting inefficiencies.
Key takeaways emphasize the growing use of context anchoring to mitigate AI memory limitations, the resurgence of CLIs and decentralized web practices, and the need for mindful AI adoption. The episode also touches on cultural nostalgia for pre-web 2.0 digital creativity while cautioning against the risks of overhyping AI tools. A recurring theme is the balance between leveraging AI for productivity and maintaining human accountability, transparency, and alignment with organizational goals.
17 Nov 2026 You cant blame AI for the code you ship | Freshworks Murali Swaminathan
Freshworks uses an AI Product Development Lifecycle to manage code quality and prioritize measurable customer impact over mere tool adoption.
3 Nov 2026 Bet on research, not on vibes | Together AIs Dan Fu
Domain expertise amplifies AI's impact on GPU kernel development, while organizations must adapt management practices to attribute costs and distribute individual engineering gains.
13 Oct 2026 Building the software factory from the spec up | LinearB's Ariel Illouz & Ofer Affias
"Linear B adopted AI-assisted, spec-driven development, starting small and scaling through structured practices, AI guilds, and iterative refinement, while addressing challenges in code context and engineer role shifts amid rapid industry changes."
29 Sept 2026 Why AI made code review the new bottleneck, and the metric that spotted it | Newsela's Dee Wilcox
"AI's impact on engineering teams is examined, revealing faster development cycles but longer reviews, requiring workflow adjustments, KPI tracking, and iterative testing, while balancing innovation, governance, and quality assurance."
15 Sept 2026 Treat your data pipelines like code, because your agents already do | MotherDuck's Jordan Tigani
"Modern data infrastructure evolves with AI-driven automation, smaller efficient systems, and tools like MotherDuck, balancing innovation, cost, and human collaboration."