Agentic Engineering Mirrors First-Principles Medical Innovation - Episode Hero Image

Agentic Engineering Mirrors First-Principles Medical Innovation

The Changelog: Software Development, Open Source · · Listen to Original Episode →
Original Title:

TL;DR

  • Agentic coding platforms like Kilo Code enable developers to achieve output speeds previously requiring teams of seven, by leveraging AI agents to parallelize tasks and accelerate development cycles.
  • The rapid evolution of AI models, with new free and low-cost options emerging faster than paid ones are deprecated, allows Kilo Code to offer users extensive token usage at list price.
  • Kilo Code's "all-in-one" approach, integrating coding, review, security, and deployment across multiple platforms, aims to replicate GitLab's success by eliminating application switching for developers.
  • The future of developer interaction with AI will shift from IDE-centric interfaces to more conversational, Slack-like experiences, enabling seamless collaboration with AI agents.
  • Agentic coding platforms are democratizing software creation, reducing the need for formal computer science degrees and tedious coding challenges, allowing individuals to build complex software faster.
  • Kilo Code's open-core model, offering paid features for managerial insights and enterprise needs like centralized key management, balances accessibility with revenue generation.
  • The increasing capability of AI models, approaching human-level intelligence, signifies a potential paradigm shift where knowledge workers can achieve significantly higher productivity with AI assistance.

Deep Dive

Sid Sijbrandij's journey, marked by founding GitLab and a personal battle with bone cancer, underscores a profound shift in software development and healthcare. His experience highlights how proactive, first-principles thinking can drive innovation in both technology and medicine, leading to the development of new companies like Kilo Code, which aims to revolutionize agentic engineering by mirroring this parallel processing approach.

Sijbrandij's personal health crisis catalyzed a deep dive into experimental treatments, demonstrating an "and then what?" approach to medical care. Faced with limited standard options, he pursued parallel treatment strategies, combining drugs and diagnostics from various sources. This necessity-driven innovation mirrors his approach to building technology companies. The implication is that a rigid, sequential approach to problem-solving, common in traditional systems, can be a liability when facing complex, time-sensitive challenges. His pursuit of treatments, including traveling to China for advanced diagnostics and using experimental therapies, illustrates a willingness to challenge established protocols for the sake of efficacy. This mindset directly informs Kilo Code's architecture, which emphasizes parallel processing and the integration of diverse AI models.

The core argument is that a "first-principles" and "parallel processing" mindset, honed in software development, is equally applicable and often more effective in complex fields like medicine. Sijbrandij's experience with cancer treatments, where he combined multiple therapies and used advanced diagnostics like single-cell sequencing, directly translates to Kilo Code's strategy of supporting a vast array of AI models and enabling parallel agentic workflows. The implication is that innovation often arises not from incremental improvements within existing frameworks, but from challenging those frameworks and exploring novel combinations. His proactive approach to his own health, treating it as a problem to be solved with maximum possible options, has led to a therapeutic "ladder" and the development of new drugs. This mirrors his strategy with Kilo Code, where he aims to provide an all-in-one, open-source platform that integrates various AI capabilities, anticipating the future of software development as a collaborative effort between humans and intelligent agents.

The second-order implication is that this paradigm shift--from sequential to parallel, from single solutions to integrated platforms--will redefine not only software engineering but also potentially accelerate progress in other fields, particularly medicine. Sijbrandij's personal journey has become a testing ground for these principles, demonstrating that a proactive, multi-pronged approach, driven by a deep understanding of underlying mechanisms and a willingness to integrate diverse tools, can yield significant breakthroughs. His ongoing development of ten different drugs and diagnostics, alongside founding Kilo Code, signifies a belief that these integrated, parallel approaches are the future, offering a more effective way to tackle complex challenges.

Action Items

  • Audit build pipeline: Identify 5 slowest steps and establish 10-minute CI target to maintain fast feedback.
  • Create runbook template: Define 5 required sections (setup, common failures, rollback, monitoring) to prevent knowledge silos.
  • Implement mutation testing: Target 3 core modules to identify untested edge cases beyond coverage metrics.
  • Track 5-10 high-variance events per game (fumble recoveries, special teams plays) to measure outcome impact.
  • Measure team strength disconnect: For 3-5 teams, calculate correlation between win-loss record and power ranking score.

Related Episodes

Key Quotes

"we tried to lead the company with a ton of transparency and we believed that helped us be an all remote company but at the same time be super on the same page and it it paid its dividends both in coordinating us hiring great people but also getting customers excited about the company itself and and how we would partner"

Sid Sijbrandij explains that GitLab's commitment to transparency was instrumental in fostering an all-remote culture while ensuring team alignment. This approach not only facilitated hiring top talent but also generated enthusiasm among customers regarding the company's vision and collaborative potential.

"end of 22 one year after the ipo i discovered i had bone cancer a 6 centimeter tumor grown from my spine and we had to do surgery really quickly i was in a lot of pain we removed the vertebrae and most of the cancer and then did a spinal fusion with a titanium frame we did radiation we did chemo and i did my first single patient ind"

Sid Sijbrandij recounts the critical juncture of discovering bone cancer shortly after GitLab's IPO, detailing the immediate surgical intervention, subsequent treatments like radiation and chemotherapy, and his participation in a single-patient investigational new drug (IND) trial. This period marked a significant personal health crisis that necessitated a shift in his focus.

"if you want to be able to reproduce you need to isolate the change so when i started doing parallel treatments one doctor told me two of my face but what if it works we won't know what would have cured you and then i let that doctor know that i wasn't interested in finding what cured me i was interested in getting cured"

Sid Sijbrandij articulates his pragmatic approach to cancer treatment, prioritizing recovery over scientific reproducibility. He explains that while doctors may seek to isolate variables for research purposes, his personal goal was simply to be cured, even if it meant pursuing parallel treatments without definitively identifying which single treatment was responsible for success.

"what we saw was that the t cell infiltration went up a lot it went up from 20 to 90 so it was the place was buzzing with t cells which is a good thing and we think that is a combination of the two checkpoint inhibitors that i'm taking that kind of unleash your immune system and one experimental treatment that is an oncolytic virus so like a modified common cold virus that drops the tgf beta and the tgf beta is something the cancer uses to hide from the immune system and it kind of tears away that invisibility cloak"

Sid Sijbrandij describes a positive development in his cancer treatment, noting a significant increase in T-cell infiltration, which indicates a heightened immune response. He attributes this to a combination of checkpoint inhibitor drugs and an oncolytic virus therapy designed to neutralize a protein that cancer cells use for immune evasion, effectively removing their "invisibility cloak."

"i think the the kind of the leading coding model that's an incredibly competitive space where kind of it always feels a bit like openai has a better model and then then google has a better model and then anthropic comes back with like the best model by far by for sure but also the best kind of free model is incredibly competitive with deep seek minimax glm all these providers and um kind of one upping each other"

Sid Sijbrandij characterizes the competitive landscape of AI coding models as dynamic and intense. He observes a continuous cycle where leading companies like OpenAI, Google, and Anthropic release advanced models, with other providers like DeepSeek, Minimax, and GLM also offering strong, often free, alternatives, creating a constantly evolving market.

"i think gentic development is another step towards this is no longer tedious i could just have an idea in my brain of how something should look how something should work and i can just write that down or even use voice mode in kilo and it will it will get done it's amazing like it's never been a better time to to create software because we can do so much more so much quicker and it doesn't have to take you 10 years anymore you don't have to study computer science you don't have to solve these lead coding problems anymore you can just make it yourself without any other human helping you it's incredible"

Sid Sijbrandij expresses optimism about the future of software development, viewing agentic development as a significant advancement that reduces tedium. He believes that individuals can now translate ideas directly into functional software, either through text or voice commands, making it faster and more accessible to create software without extensive formal training or traditional coding challenges.

Resources

External Resources

Books

  • "The American Dream" - Mentioned in relation to Sid's personal journey and success.

Articles & Papers

  • "Stop Collaboration" (Post Hard Founders) - Referenced as an article advocating for individual progress over excessive collaboration.

People

  • Sid Sibrandidge - Founder of GitLab and Kilo Code, discussed his entrepreneurial journey, health battle, and ventures.
  • Cal Goldbreth - Co-founder and CEO of Depot Dev, discussed build optimization.
  • Tyler Cowen - Mentioned in relation to the concept of Artificial General Intelligence (AGI) in April.
  • Matt German - CEO of AWS, mentioned regarding the role of junior developers and AI.
  • Coleson Broda - Mentioned as an example of a polymath.
  • PC - Mentioned as an example of a polymath.
  • Breakmaster Cylinder - Provided music for the podcast.
  • Matt Ray - Host of "changelog and friends" podcast.

Organizations & Institutions

  • GitLab - Mentioned as the company Sid Sibrandidge led to IPO, and for its ongoing observability project.
  • Kilo Code - Mentioned as an agentic engineering platform founded by Sid Sibrandidge.
  • Depot Dev - Discussed for its role in optimizing build times.
  • Amazon - Provider of ARM and AMD CPUs used by Depot Dev.
  • GitHub - Mentioned in relation to its hosted runners and cache multiplexing.
  • Y Combinator - Mentioned as the incubator for a company Sid invested in.
  • Open Core Ventures - Mentioned as a company Sid is creating around open-source projects.
  • OpenAI - Mentioned in relation to its models and as a user of Notion.
  • Anthropic - Mentioned in relation to its AI models.
  • Google - Mentioned in relation to its AI models, including Gemini Flash.
  • Tiger Data - Provider of agentic PostgreSQL database solutions.
  • Ramp - A company that uses Notion.
  • Vercel - A company that uses Notion.
  • Cloudflare - Provider of containers used for Kilo Code's cloud agents.
  • AWS (Amazon Web Services) - Mentioned in relation to CEO Matt German's comments.
  • Post Hard Founders - Authors of an article on collaboration.

Tools & Software

  • Fly io - Platform for developers, mentioned as a partner.
  • GitHub Actions - Discussed for cache multiplexing within Depot Dev's runners.
  • Kilo (CLI) - Command-line interface for Kilo Code.
  • Kilo (Mobile App) - Upcoming mobile application for Kilo Code.
  • Kilo (Cloud Agents) - Cloud-based agents for Kilo Code.
  • Kilo (App Builder) - Web-based experience for Kilo Code.
  • Notion - Productivity and note-taking tool with AI features.
  • Notion Agent - AI assistant within Notion.
  • Cursor - AI-powered code editor, compared to Kilo.
  • VS Code - Code editor where Kilo is available.
  • JetBrains - Code editor suite where Kilo is available.
  • Slack - Communication platform, mentioned for Kilo's integration.

Websites & Online Resources

  • changelog.com - Website for the podcast.
  • changelog.com/request - Webpage for submitting podcast requests.
  • tigerdata.com - Website for Tiger Data.
  • notion.com/changelog - Specific URL for trying Notion with the podcast's link.

Other Resources

  • Agentic Engineering Platform - Category of platform Kilo Code belongs to.
  • Artificial General Intelligence (AGI) - Discussed as models becoming as smart as the median human.
  • Click Chemistry - A method for combining drugs with binders.
  • FAP Binder - A binder that targets fibroblast.
  • HLA Type - A genetic factor that disqualified Sid from some trials.
  • Osteosarcoma - The type of bone cancer Sid has.
  • TFE3 Amplification - A genetic anomaly found in Sid's cancer.
  • Fibroblast - Scar tissue mentioned in relation to cancer treatment.
  • Lutetium - A radioactive element used in treatment.
  • Ettrium - A radioactive element used in treatment.
  • TGF Beta - A substance cancer uses to hide from the immune system.
  • B7H3 Expression - A marker found to be higher in Sid's liver.
  • Single Cell Sequencing - A technology used to analyze tumor cells.
  • Checkpoint Inhibitors - A type of immune system treatment.
  • Oncolytic Virus - A modified virus used in cancer treatment.
  • mRNA Vaccine - A type of vaccine with potential curative properties.
  • T Cell Infiltration - An indicator of immune system activity against cancer.
  • Histochemistry - A technique for coloring tissue slides.
  • PET Scan - A medical imaging technique.
  • CT Scan - Mentioned in relation to diagnostic imaging.
  • HLA Type - A genetic factor that disqualified Sid from some trials.
  • Osteosarcoma - The type of bone cancer Sid has.
  • TFE3 Amplification - A genetic anomaly found in Sid's cancer.
  • Fibroblast - Scar tissue mentioned in relation to cancer treatment.
  • Lutetium - A radioactive element used in treatment.
  • Ettrium - A radioactive element used in treatment.
  • TGF Beta - A substance cancer uses to hide from the immune system.
  • B7H3 Expression - A marker found to be higher in Sid's liver.
  • Single Cell Sequencing - A technology used to analyze tumor cells.
  • Checkpoint Inhibitors - A type of immune system treatment.
  • Oncolytic Virus - A modified virus used in cancer treatment.
  • mRNA Vaccine - A type of vaccine with potential curative properties.
  • T Cell Infiltration - An indicator of immune system activity against cancer.
  • Histochemistry - A technique for coloring tissue slides.
  • PET Scan - A medical imaging technique.
  • CT Scan - Mentioned in relation to diagnostic imaging.
  • MDM2 - A gene mentioned in relation to cancer.
  • P53 Division - A cellular process related to MDM2.
  • TCRT - A type of medical treatment.
  • CART - A type of medical treatment.
  • Rubi - A programming language mentioned for its efficiency.
  • Flow State - A psychological state of deep engagement.
  • Agentic Coding - A method of software development using AI agents.
  • Open Source - A development model discussed for Kilo Code.
  • Open Core - A business model combining open source with paid features.
  • Bring Your Own Keys (BYOK) - An enterprise feature for security.
  • VPC (Virtual Private Cloud) - A security feature for enterprise.
  • Grok - An AI model.
  • Minimax - An AI model.
  • GLM - An AI model.
  • Deepseek - An AI model.
  • Gemini Flash - An AI model from Google.
  • GPT-5 - A future AI model from OpenAI.
  • Opus 45 - An AI model.
  • Jira - Project management tool.
  • Linear - Project management tool.
  • Java - Programming language.
  • Python - Programming language.
  • Go CLI - A command-line interface tool.
  • CI (Continuous Integration) - A software development practice.
  • CD (Continuous Deployment) - A software development practice.
  • Observability - A concept in software development.
  • Vector Database - A type of database for AI.
  • Elasticsearch - A search and analytics engine.
  • PostgreSQL - A relational database.
  • MCP (Model Context Protocol) - A protocol for AI agents.
  • Hybrid Search - A search method combining vector and keyword search.
  • Zero Copy Forks - A database feature for efficient cloning.
  • Copy on Write - A data management technique.
  • Time Series - Data collected over time.
  • Embeddings - Numerical representations of data.
  • Relational Data - Data organized in tables.
  • Conversational History - Records of conversations.
  • Plutonium - A radioactive element, mentioned in an anecdote.
  • T-cell - A type of white blood cell.

---
Handpicked links, AI-assisted summaries. Human judgment, machine efficiency.
This content is a personally curated review and synopsis derived from the original podcast episode.