Agentic Architecture: Why Files Aren't Always Enough
<p>What are the limitations of using a file-based agent workflow? Why do massive context windows tend to collapse? This week on the show, Mikiko Bazeley from MongoDB joins us to discuss agentic architecture and context engineering.</p>
<p>Mikiko is an applied AI engineer. She helps developers and organizations build AI and ML applications using MongoDB. We dig into the debate of files versus a database. What are some of the limitations of building an agent with just a folder of files? </p>
<p>We explore the surprising limitations of massive context windows and strategies for fixing them. Mikiko also shares advice and resources to help you get up to speed on building your own agent skills. Our conversation touches on multiple topics in the current development landscape.</p>
<p>This episode is sponsored by <a href="https://serpapi.com/?utm_source=realpython&utm_medium=podcast&utm_campaign=q12">SerpApi</a>.</p>
<div class="alert alert-primary" role="alert">
<p><strong>Video Course Spotlight:</strong> <a href="https://realpython.com/courses/building-type-safe-llm-agents-with-pydantic-ai/">Building Type-Safe LLM Agents With Pydantic AI</a></p>
<p>Build type-safe LLM agents in Python with Pydantic AI using structured outputs, function calling, and dependency injection.</p>
</div>
<p>Topics:</p>
<ul>
<li>00:00:00 – Introduction</li>
<li>00:02:31 – Catching up with MongoDB</li>
<li>00:07:02 – Are the files all you need?</li>
<li>00:15:14 – What is a workflow agent?</li>
<li>00:24:43 – Sponsor: SerpApi</li>
<li>00:25:45 – Model vs harness</li>
<li>00:29:57 – Context rot and tool loadouts</li>
<li>00:41:07 – Sharing state and coordination of agents</li>
<li>00:47:27 – Video Course Spotlight</li>
<li>00:49:16 – What do dataflows look like</li>
<li>01:00:38 – The human-in-the-loop & coding agents</li>
<li>01:10:30 – Resources to explore</li>
<li>01:17:49 – What are you excited about in the world of Python?</li>
<li>01:18:38 – What do you want to learn next?</li>
<li>01:22:54 – Thanks and goodbye</li>
</ul>
<p>Show Links:</p>
<ul>
<li><a href="https://thenewstack.io/ai-agent-memory-architecture/">The “files are all you need” debate misses what’s actually happening in agent memory architecture - The New Stack</a></li>
<li><a href="https://www.mongodb.com/">MongoDB: The World’s Leading Modern Data Platform</a></li>
<li><a href="https://venturebeat.com/data/karpathy-shares-llm-knowledge-base-architecture-that-bypasses-rag-with-an">Karpathy shares ‘LLM Knowledge Base’ architecture that bypasses RAG with an evolving markdown library maintained by AI - VentureBeat</a></li>
<li><a href="https://www.llamaindex.ai/blog/files-are-all-you-need">Files Are All You Need: Context, Search, Skills Guide | LlamaIndex</a></li>
<li><a href="https://www.mongodb.com/company/blog/technical/converged-datastore-for-agentic-ai">Converged Datastore For Agentic AI - MongoDB</a></li>
<li><a href="https://thenewstack.io/why-developers-need-vector-search/">Why Developers Need Vector Search - The New Stack</a></li>
<li><a href="https://www.oreilly.com/radar/why-multi-agent-systems-need-memory-engineering/">Why Multi-Agent Systems Need Memory Engineering – O’Reilly</a></li>
<li><a href="https://www.philschmid.de/context-engineering">The New Skill in AI is Not Prompting, It’s Context Engineering - Phil Schmid</a></li>
<li><a href="https://www.dbreunig.com/2025/06/22/how-contexts-fail-and-how-to-fix-them.html">How Long Contexts Fail - dbreunig.com</a></li>
<li><a href="https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html">How to Fix Your Context - dbreunig.com</a></li>
<li><a href="https://arxiv.org/abs/2601.11653">AI Agents Need Memory Control Over More Context - arxiv.org</a></li>
<li><a href="https://www.latent.space/p/ainews-is-harness-engineering-real">AINews - Is Harness Engineering real? - Latent.Space</a></li>
<li><a href="https://adambaitch.substack.com/p/the-model-vs-the-harness-which-actually-matters-more-59dd3116bb31">The Model vs. the Harness: Which Actually Matters More?</a></li>
<li><a href="https://realpython.com/chromadb-vector-database/">Embeddings and Vector Databases With ChromaDB – Real Python</a></li>
<li><a href="https://github.com/humanlayer/12-factor-agents">12-factor-agents: What are the principles we can use to build LLM-powered software that is actually good enough to put in the hands of production customers?</a></li>
<li><a href="https://12factor.net/">The Twelve-Factor App</a></li>
<li><a href="https://learn.mongodb.com/">MongoDB Courses and Trainings - MongoDB University</a></li>
<li><a href="https://github.com/mongodb-js/mongodb-mcp-server">mongodb-mcp-server: A Model Context Protocol server to connect to MongoDB databases and MongoDB Atlas Clusters.</a></li>
<li><a href="https://www.mongodb.com/docs/mcp-server/">What is the MongoDB MCP Server? - MongoDB Docs</a></li>
<li><a href="https://github.com/mongodb/mongo-python-driver">mongo-python-driver: PyMongo - the Official MongoDB Python driver</a></li>
<li><a href="https://github.com/mongodb/agent-skills">agent-skills: Use the official MongoDB Skills with your favorite coding agent to build faster.</a></li>
<li><a href="https://reachymini.net/">Reachy Mini - Open-Source Desktop Humanoid Robot</a></li>
<li><a href="https://www.linkedin.com/in/mikikobazeley/">👩🏻💻 Mikiko B. - LinkedIn</a></li>
<li><a href="https://mikikobazeley.substack.com/">Building AI Products From Scratch - Mikiko Bazeley - Substack</a></li>
</ul>
<p>Level up your Python skills with our expert-led courses:</p>
<ul>
<li><a href="https://realpython.com/courses/getting-started-claude-code/">Getting Started With Claude Code</a></li>
<li><a href="https://realpython.com/courses/building-type-safe-llm-agents-with-pydantic-ai/">Building Type-Safe LLM Agents With Pydantic AI</a></li>
<li><a href="https://realpython.com/courses/pydantic-simplify-data-validation/">Using Pydantic to Simplify Python Data Validation</a></li>
</ul> <p><a rel="payment" href="https://realpython.com/join">Support the podcast & join our community of Pythonistas</a></p>
The Grittiest Conversations of 2025: AI, Business & Beyond
In this recap episode, we highlight the best moments from our 2025 interviews and reflect on the ideas that defined the year.
Featuring:
David Rubenstein (co-founder of Carlyle)
Yamini Rangan (CEO of HubSpot)
Ben Chestnut (co-founder of Mailchimp)
Winston Weinberg (co-founder and CEO of Harvey)
Garrett Lord (co-founder of Handshake)
Aidan Gomez (co-founder and CEO of Cohere)
Michelle Zatlyn (co-founder of Cloudflare)
Evan Spiegel (co-founder and CEO of Snap)
Connect with Joubin
X: https://x.com/Joubinmir
LinkedIn: https://www.linkedin.com/in/joubin-mirzadegan-66186854/
Email: grit@kleinerperkins.com
Follow on LinkedIn:
https://www.linkedin.com/company/kpgrit
Follow on X:
https://x.com/KPGrit
Learn more about Kleiner Perkins: https://www.kleinerperkins.com/
(July 2021) Ben Chestnut: From Side Business to $12 Billion – The Accidental Triumph of Mailchimp
In 2001, three web designers built a quirky email tool called Mailchimp.
It wasn’t their main business.
It wasn’t even meant to make real money.
For years, Ben Chestnut and his partners survived on web-design gigs, while Mailchimp earned just enough to cover lunch.
Then…one surprising spreadsheet changed everything.
In this episode, Ben reveals:
The decision that tripled their revenue overnight
How a mispronunciation on the world's biggest podcast turned into a marketing masterstroke
How a painful wake-up call about leadership reshaped the company’s culture
Listen now to hear how guerrilla billboards, a monkey logo, and a string of improbable breaks built one of the most recognizable names in tech.
Follow How I Built This:
Instagram → @howibuiltthis
X → @HowIBuiltThis
Facebook → How I Built This
Follow Guy Raz:
Instagram → @guy.raz
X → @guyraz
Substack → guyraz.substack.com
Website → guyraz.com
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
Bootstrapped to $12B: Mailchimp’s Ben Chestnut on Life After the Exit
Guest: Ben Chestnut, Former CEO and Co-Founder of Mailchimp
If you find yourself selling your startup, then Mailchimp co-founder Ben Chestnut has some important advice for you: Get a dog.
When Intuit bought Mailchimp in 2021 for $12 billion, the company asked Ben if he wanted to stay on as CEO, but he chose to “walk off into the sunset” and let the new owners take over.
After that, he estimates it took 6 to 12 months before he stopped checking his email, social media, and calendar with the same level of stress a CEO might have. Adopting a dog, he discovered, forces you to “get OK with the voices in your head."
“After the acquisition, that's all I do, I walk the dog,” Ben says. “And the dog was good therapy ... No judgments from a dog.”
Mentioned in this episode: Intuit, Wolt, DoorDash, LinkedIn, Dan Kurzius, Salesforce, ExactTarget, Pardot, Constant Contact, Rackspace, Free by Chris Anderson, Wired Magazine, Charles Hudson, the Freemium Summit, Drew Houston, Dropbox, Evernote, Phil Libin, TechCrunch, Brian Kane, Catalyst Partners, Georgia Pacific, Scott Cook, Bing Gordon, Vinay Hiremath, Loom, Joe Thomas, Caltrain, Flickr, Saturday Night Live, Droga5, Cannes Film Festival, Strava, Twitter, LinkedIn, Nvidia, Glean, Rubrik, Amazon AWS, and Mechnical Turk.
Links:
Connect with Ben
LinkedIn
Connect with Joubin
Twitter
LinkedIn
Email: grit@kleinerperkins.com
Learn more about Kleiner Perkins
20VC: Mailchimp's Ben Chestnut on The Biggest Leadership Lessons Scaling to a $12Bn Acquisition and $1BN+ ARR, The Secret to Happiness, Being a Great Husband and Father & Why 2021 was the Right Time To Sell to Intuit
Ben Chestnut is the Co-Founder of Mailchimp, the all-in-one marketing platform for small businesses. Last year, in Sept 2021 it was announced that Intuit would acquire Mailchimp for a reported $12BN. There are so many things to love about the Mailchimp journey to this point. First, Mailchimp was founded as the result of a side project of a design agency Ben and his co-founder, Dan, used to run. Second, Mailchimp is and has always been based in Atalanta, eschewing the notion you have to be in SF or NYC to build a massive business. Then third, they never raised venture funding for the business all the way until their $12BN acquisition. Ben led Mailchimp to over 1,200 employees and millions of global users.
In Today's Episode with Ben Chestnut We Discuss:
1. From Mama's Kitchen to the Smell of Business and Founding Mailchimp:
How did Ben turn a mediocre agency into the founding of Mailchimp? What was the a-ha moment?
At what stage of the business did Ben quit the agency and go all in on Mailchimp? What sign did he need that Mailchimp had true product-market fit?
When Ben's mother died, he bought every flower in the local town to commemorate her. How did Ben's mother impact the type of father and husband he is today? How did she impact the way that he led Mailchimp as CEO?
Ben's fishing trips with his father played a big role in his early years, what were the single biggest lessons for Ben from his fishing trips with his father?
2. Ben Chestnut: The Leader:
How does Ben define the term "high performance" in leadership?
What does Ben mean when he says "the secret to happiness is to stay in your lane"?
Why would Ben describe himself as the "leader of the misfits"? How did that early experience and labeling impact both the people he hired and the culture he created at Mailchimp?
What does Ben mean when he says he used to have a "hands off, eyes off" leadership style? What have been the single biggest drivers in his development as a leader?
3. Ben Chestnut: The Person:
Relationship to Money: How does Ben reflect on his relationship to money? How has it changed over time? Why does Ben still to this day buy lottery tickets with his wife?
Conquering Fatherhood: What does being a great father to Ben mean? How does Ben attempt to instil the same work ethic and drive when his children are born into immense wealth?
The secret to Marriage: What does Ben believe is the core to a successful and thriving marriage? How does Ben view his role in the marriage? How has it changed over time?
Potential Lost Identity: A founder's identity is so closely tied to their company, how did Ben manage the challenge of selling his company but retaining his identity? What did Ben learn about himself through many different acquisition processes?
4. Mailchimp: The Business:
Why did Ben never raise venture money in the 21 year journey of Mailchimp?
Why did Ben never accept any of the acquisition offers that came before Intuit?
How did Ben motivate his team after they knew each acquisition offer was being turned down?
Why did Ben decide the acquisition by Intuit was the right decision for the company?
How does Ben view his role in the company now and moving forward?
SaaStr 592: Everything that Breaks on the Way to $1B ARR with Mailchimp Co-Founder Ben Chestnut and SaaStr CEO Jason Lemkin
Join Mailchimp Co-Founder Ben Chestnut and SaaStr CEO Jason Lemkin in a follow-up to their popular fireside chat held during SaaStr @ Home 2020. Ben and Jason will chat about the highlights–and the challenges–of scaling a company like Mailchimp to $1B ARR.
Full video: https://youtu.be/5fx3qFD62ak
Want to join the SaaStr community? We're the 🌎largest community for B2B software.
Subscribe for weekly updates: https://www.saastr.com/subscribeform
Twitter: https://twitter.com/saastr
LinkedIn: https://www.linkedin.com/company/2724976
Quora Group: https://www.quora.com/q/cloud
Facebook: https://www.facebook.com/SaaStr/
Instagram: https://www.instagram.com/saastr/
Our North American Event: https://bit.ly/2OXeAYh
Our European Event: https://bit.ly/2OZTad8
599: MLOps: Machine Learning Operations
This week, Mikiko Bazeley, Senior Software Engineer at Mailchimp joins the podcast to share her in-depth knowledge of MLOps: Machine Learning Operations. Tune in to hear her discuss what it entails, why it's so critical for the efficiency of any data science team, and the most important tools you need to master for career success in this field.
In this episode you will learn:
• What MLOps is [11:40]
• Mikiko’s role at Mailchimp and why MLOps is critical for the efficiency of any data science team [27:11]
• The three most important MLOps tools [32:15]
• The six most essential MLOps skills for data scientists [47:01]
• The key factors Mikiko looks when hiring engineers [1:07:31]
• Mikiko’s productivity tricks for balancing software engineering, content creation, and her athletic pursuits [1:13:20]
Additional materials: www.superdatascience.com/599
The case for bootstrapping, w/Mailchimp's Ben Chestnut
This week, we’re returning to an episode of Masters of Scale featuring Ben Chestnut, the co-founder and CEO of Mailchimp. The email marketing platform was acquired this year by Intuit in a multi-billion-dollar deal that closed this month. Hear how Chestnut used a DIY ethos to grow the company without ever raising a dollar of outside funding. The Mailchimp story is the exception to Reid's rule (always raise more money than you think you need!). The episode explores a range of options for those who don't fit the VC-funding mold. Cameo appearances: LeVar Burton (Star Trek, Reading Rainbow, LeVar Burton Reads), Don MacKinnon (Milq), Karen Cahn (iFundWomen).
Read a transcript of this episode: https://mastersofscale.com
Subscribe to the Masters of Scale weekly newsletter: https://mastersofscale.com/subscribe
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
Mailchimp: Ben Chestnut
In the late 1990s, Ben Chestnut was a struggling young designer interning at an appliance company, when somebody suggested that he try designing for the internet instead. A few years later, Ben and two co-founders launched a web design agency, only to discover that the service they'd included almost as an afterthought—email marketing—was taking off among their small-business clients. The founders named that service Mailchimp and pivoted to it full-time in 2007, choosing a winking monkey as their mascot, and stumbling onto the Freemium model before it became mainstream. But their most impeccable timing came in 2014, when they decided to sponsor a new podcast called Serial, a move that catapulted the winking monkey into popular culture. Over the years, despite management jitters and a public reckoning over office culture, Mailchimp has remained profitable and self-funded, with revenue of $800 million in 2020.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
SaaStr 371: Learning from the Lows, How Mailchimp Navigated Economic Uncertainty with Ben Chestnut, CEO @ Mailchimp and Jason Lemkin, CEO and Founder @ SaaStr
It's no secret that businesses today are struggling with an unpredictable economy. In this session, Mailchimp Co-founder and CEO Ben Chestnut will share the story of Mailchimp's founding amidst—and despite—the dot-com bubble burst, and how the company navigated a number of inflection points in the first 10 years of its founding.
This episode is an excerpt from Jason and Ben's session at SaaStr Annual @ Home.
Full video: https://youtu.be/LfLZzakn2r0
Podcast transcript: https://www.saastr.com/saastr-podcasts-for-the-week-with-jennifer-tejada-ben-chestnut-and-jason-lemkin/
The case for bootstrapping, w/Mailchimp's Ben Chestnut
You can bootstrap your business to scale, but you'll have to make your own luck. Nobody knows this better than Mailchimp's Ben Chestnut. He used a DIY ethos to grow a $600M company without ever raising a dollar of outside funding. The Mailchimp story is the exception to Reid's rule (always raise more money than you think you need!). The episode explores a range of options for those who don't fit the VC-funding mold. Cameo appearances: LeVar Burton (Star Trek, Reading Rainbow, LeVar Burton Reads), Don MacKinnon (Milq), Karen Cahn (iFundWomen).
Read a transcript of this interview at: https://mastersofscale.com/ben-chestnut-the-case-for-bootstrapping/
Subscribe to the Masters of Scale weekly newsletter: https://mastersofscale.com/subscribe
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
.jpg?width=400){kind=small}
.jpg?width=400){kind=small}
