Lenny’s Podcast116 min
Kevin Weil · Apr 10, 2025
Key learnings- Build products at the edge of current model capabilities, since models improve every few months and a barely-working product can…
- Treat evals as unit tests for models: define hero use cases, write evals for them, and hill-climb on those evals to know whether…
- Expect AI behavior to be fuzzy: a model right 60%, 95%, or 99.5% of the time implies very different product designs, so study…
Lenny’s Podcast128 min
Nick Turley · Aug 9, 2025
Key learnings- Ship early to learn what people actually want; with AI, the valuable behaviors are emergent, so you won't know what to polish…
- Treat the model itself as a product: iterate on it using real user use cases, personality/vibe checks, and new capabilities like…
- Set the team's pace deliberately by asking whether work is 'maximally accelerated', separating fast product iteration from…
Lenny’s Podcast93 min
Mike Krieger · Jun 5, 2025
Key learnings- Use AI as an independent strategy critic: ask it to be brutal or to challenge your thinking, since default prompts tend to…
- As AI writes most code, bottlenecks shift upstream to deciding what to build and aligning people, and downstream to merge queues…
- Give teams a minimum viable strategy so people feel empowered to build and explore at the edge of model capabilities without…
Lenny’s Podcast153 min
Marc Andreessen · Jan 29, 2026
Key learnings- Expect AI to make already-skilled people far more capable; the gains are largest for those who pair deep expertise with AI tools.
- Rather than fearing job loss, focus on task loss: jobs persist while the individual tasks composing them shift, which is how…
- Combine at least two or three domains deeply (e.g., coding, product, design); the combination of skills creates far more value…
Lenny’s Podcast128 min
Dan Shipper · Jul 17, 2025
Key learnings- Non-programmers can use Claude Code-style command-line agents with local files to process large text sets, like meeting notes or…
- Hire or assign someone to an AI operations role who continuously turns repetitive team tasks into prompts and workflows, and make…
- Apply 'compounding engineering': spend a bit of effort turning each recurring task (like writing PRDs) into a reusable prompt or…
Lenny’s Podcast109 min
Benedict Evans · May 31, 2026
Key learnings- Treat AI as roughly as big as the internet or mobile, and recognize we are early, like 1997, with most applications not yet built…
- Expect adoption to be very uneven; survey data shows many people are not using AI at all, so don't assume the people in your…
- Ask whether a job is a task that can be automated or a deeper outcome; consultants and professionals are hired for judgment…
Lenny’s Podcast89 min
Michael Truell · May 1, 2025
Key learnings- Build the product you and your team actually use every day (dogfooding) so you never ship anything that isn't genuinely useful…
- When working with AI coding tools, chop tasks into small specify-generate-review loops rather than handing over one giant…
- Spend time on side projects deliberately pushing AI to its limits to build a calibrated gut feeling for what current models can…
Lenny’s Podcast111 min
Bret Taylor · Jul 31, 2025
Key learnings- Don't just digitize what existed before; rebuild the experience natively for the new platform so customers have a reason to…
- Ask yourself what the most impactful thing is today, and reflect on whether you're choosing it out of comfort or because it's…
- Be skeptical of your own storytelling: customer reasons given to salespeople can mask deeper product problems, so seek…
Lenny’s Podcast64 min
Marily Nika · Feb 5, 2023
Key learnings- Avoid the shiny object trap: only add AI when there is a real, well-defined pain point worth solving, and identify the problem…
- Don't use AI for an MVP; fake the AI behavior with a clickable prototype and test desirability with users first.
- Use ChatGPT as a sounding board for sharpening mission statements and generating user segments, personas, and motivations, while…
Lenny’s Podcast79 min
Anton Osika · Mar 9, 2025
Key learnings- Be precise and explicit when prompting an AI builder: say exactly what you expect and which parts are not working, rather than…
- Use the agent's chat mode to ask how something works or why it isn't doing what you want; this builds understanding while you…
- Expect to spend a full week taking one real problem from idea to a working product that people use; Anton suggests that alone…
Lenny’s Podcast100 min
Jenny Wen · Mar 1, 2026
Key learnings- Classic linear design process (research, diverge, converge) is breaking down as engineers ship prototypes quickly; designers…
- Design work is splitting into two modes: supporting implementation and setting a vision, where the vision now often spans three…
- Designers can use coding tools to do last-mile polish and prototype in real code, rather than waiting on engineers for every…
Lenny’s Podcast117 min
Sander Schulhoff · Jun 19, 2025
Key learnings- Prompt engineering remains valuable: Sander argues that bad prompts can drop accuracy near zero on a task while good prompts can…
- Few-shot prompting is the highest-impact basic technique: give the model several examples of what good output looks like, using a…
- Ask the model to decompose a problem by first listing the subproblems it needs to solve, then solving each one before answering…
Lenny’s Podcast134 min
Claire Vo · Mar 29, 2026
Key learnings- Run several purpose-built agents instead of one general agent; splitting tasks keeps each agent's context focused and reduces…
- Install OpenClaw on a separate clean machine such as an old laptop or Mac Mini, with its own local admin account, rather than…
- Onboard the agent like a new employee: give it its own email and calendar, share access selectively, and expand trust…
Lenny’s Podcast131 min
Hamel Husain & Shreya Shankar · Sep 25, 2025
Key learnings- Evals are systematic ways to measure and improve an AI application, essentially data analytics on LLM behavior, replacing…
- Start with error analysis by manually reviewing around 100 sampled traces and writing short notes on the first upstream failure…
- Product people with domain expertise should lead open coding; appoint one 'benevolent dictator' whose judgment you trust instead…
Lenny’s Podcast87 min
Elizabeth Stone · Jul 19, 2026
Key learnings- Expect a 'storming' phase with new technology; don't abandon AI, but be deliberate about capturing benefits while limiting costs…
- Let product, design and data people prototype early against a clearly defined business problem, while still working with…
- Invest in source-of-truth data, guardrails for shipping and testing, and review processes where AI output needs verification, and…
Lenny’s Podcast100 min
Zevi Arnovitz · Jan 18, 2026
Key learnings- Start slowly with a ChatGPT project acting as a skeptical 'CTO' before moving to coding tools, gradually easing into code to…
- Capture ideas quickly as Linear issues using reusable slash commands, so you can pick them up later with context already gathered.
- Separate the workflow into explore, plan, execute, review, and document phases, and have the AI ask clarifying questions before…
Lenny’s Podcast108 min
Cat Wu · Apr 23, 2026
Key learnings- Shorten the path from idea to user: AI-era timelines have shrunk from six months to days or a week, so PMs should focus on…
- Ship features as clearly labeled research previews to lower the commitment of launching and get fast user feedback that can be…
- Set a clear, specific goal for who the key user is and what success means, which rules out many approaches and speeds decisions.
Lenny’s Podcast78 min
Dr. Fei Fei Li · Nov 16, 2025
Key learnings- Large labeled datasets were the missing ingredient that unlocked modern AI; big data, neural networks, and GPUs together formed…
- Pursue a north-star problem and commit to it for years, as ImageNet's object recognition focus did for her lab.
- Be willing to take intellectual and career risks, such as leaving tenure track or joining new ventures, when the mission and…
Lenny’s Podcast97 min
Edwin Chen · Dec 7, 2025
Key learnings- Define quality in rich, specific terms for each domain (e.g., what makes a poem Nobel-worthy, not just whether it has eight…
- Gather thousands of signals on each worker and task, from background and expertise to performance, and use them like an ML…
- Don't trust benchmarks at face value: many contain wrong answers, and optimizing for them can make models worse at messy…
Lenny’s Podcast102 min
Chip Huyen · Oct 23, 2025
Key learnings- Focus on what actually improves AI apps, such as talking to users, improving data, writing better prompts, and optimizing…
- Be cautious about committing to newly released technologies that have not been widely tested, since switching away from them…
- Fine-tuning and post-training can shape model behavior a lot, and many teams now focus effort there because base pre-training…
Lenny’s Podcast123 min
Lazar Jovanovic · Feb 8, 2026
Key learnings- Start a project with several parallel attempts (brain dump, typed prompt, design reference, code template) to find the strongest…
- Spend most of your time planning and chatting with the AI rather than executing, since clarity matters more than raw build speed.
- Write a set of source-of-truth docs (master plan, implementation plan, design guidelines, user journeys, tasks.md) so the agent…
Lenny’s Podcast102 min
Dianne Penn · Jul 26, 2026
Key learnings- Treat evals as the new PRDs: turn vague user complaints into concrete, reproducible test sets that researchers can act on and…
- Dig into user transcripts, not just pixels, to find where the model actually failed, such as wrong tool calls, bad retrieval, or…
- Keep strong conviction about a theme or area while staying loose about the specific prototype, and revisit bets across one to two…
Lenny’s Podcast102 min
Paul Adams · Oct 26, 2023
Key learnings- Before redesigning a product for AI, start from the core premise: why people use it and what problem it solves, then map what AI…
- Treat AI as a strategic shift rather than a feature; Intercom ripped up its strategy and rebuilt from first principles after the…
- Build real machine learning depth on the team, but spread AI skills across product teams instead of isolating AI work in a…
Lenny’s Podcast148 min
Amol Avasare · Apr 5, 2026
Key learnings- Expect most growth time to go to 'success disasters' once growth is rapid; firefighting breaking systems matters as much as new…
- In AI-first products, skew the growth portfolio toward larger swings, since future product value could be orders of magnitude…
- Treat activation as a top lever in AI products: identify user traits that predict which feature fits them, then guide them there.