Method · MoneyAnish Acharya relays an approach from Ale at Kavak (which sells used cars online with an agent per customer): when an agent gets stuck, it calls a human, and the human coaches the agent through. The agent then captures all the traces and learns from it, so the next time it shouldn't have to call the human. It's either a knowledge gap or a data gap you have to give the agent.
Method, not a verified protocol.
Lenny's Podcast · Why companies are becoming a series of loops | Anish Acharya (a16z)
Recommendation · SocialAfter opening with intent, don't go into a name introduction — the coach calls it an additional unnecessary escalation. Instead, elaborate on what you noticed about the woman, her outfit, or what she's up to. This makes the conversation more relevant to her and is less of an escalation. He also discourages asking 'what's your name?' and 'where are you from?' back to back right after opening.
Coach Kyle · How to Talk to Women in Public: Real Infield Breakdown
Recommendation · SocialThe coach endorses a direct opener like: 'Hey, excuse me. I was just walking by and I saw you and I just thought you looked really cute. So I want to come say hi.' He notes this phrasing is fine if you need a completely canned generic opener, and that the client's use of it was done well.
Coach Kyle · How to Talk to Women in Public: Real Infield Breakdown
Recommendation · MoneyTim's learning from the outsiders-glow mistake: when building a leadership team and you notice a missing in-house skill, don't default to hiring experienced outsiders — look at internal candidates first, and give internal people the benefit of the doubt way more than outsiders. He acknowledges that betting on an internal person is a personal investment and a bigger risk on a person than hiring externally.
The speaker explicitly frames this as his personal learning; the risk trade-off is his judgment.
MicroConf · Scaling a SaaS to €35M ARR: 3 Mistakes I'd Never Repeat
Recommendation · MoneyTim's fix for the culture mistake is to write down a one-pager describing what you think the culture is — a framework is optional, but the act of writing makes a huge difference. Writing it down makes you actually think about it and gives you something to align with other people. If you're with two founders, doing the wordsmithing together will amaze you with how much you actually differ.
The claim about founders discovering differences is the speaker's expectation from experience, not a measured result.
MicroConf · Scaling a SaaS to €35M ARR: 3 Mistakes I'd Never Repeat
Method · SocialWhen asked what you do for work, say you'll answer but ask her a question first: which search engine she uses. If she says Google, respond 'Okay, that's what I'm talking about, high five' and explain you thought she'd say something weird like Yahoo or Bing; if she says another engine, tease her with 'Oh really, interesting... why not Google?' Then reveal your job with a fun backstory and a funny low-key flex line.
Method, not a verified protocol.
Coach Kyle · From Hi to Coffee Date (LIVE DEMO)
RecommendationStep one of cooking with Codex is giving Codex your context from documents, your brain, conversations, email, Notion, Linear, and so forth. Step two is to just build things once you've given it context. A large code base's Git commits and history serve as that context.
Also discussed by Jake on Figma (unclassified source — not counted until classified). Not counted as independent sources.
AI Engineer · Cooking with Codex — Charlie Guo & Gabriel Chua, OpenAI
Recommendation · MoneyKent C. Dodds recommends launching something small within weeks and testing the market manually, rather than investing a whole team for a year. He reflects that this approach would have allowed testing the market and learning very quickly without the large investment, calling project validation really important.
AI Engineer · Build the Right Thing: Product Engineering (Part 1) — Kent C. Dodds, EpicProduct.engineer
Method · SocialA step-by-step opening sequence for approaching a woman in a mall: get her attention, stop at a good distance, secure her undivided attention, then deliver the open, followed by an observation rather than a question.
Method, not a verified protocol.
Coach Kyle · A Realistic Approach at a Mall (Infield)
Method · SocialCoach Kyle models a full cold approach on a woman sitting alone: approach slowly and relaxed, vibe, tease, and flirt, sit down next to her rather than standing over her, and start a conversation, eventually asking her to get coffee.
Method, not a verified protocol.
Coach Kyle · A Realistic Approach to a Single Woman Sitting Alone on a Bench (Infield)
Recommendation · MoneyKent C. Dodds advises that if a task comes without a job to be done attached, go back to the PM and say you don't know enough about what the task is about to design a good system for it. Without the job defined, you can build a requested feature while totally missing the parts of the system or the primitives users actually need, and risk spending three months on something that doesn't hit users.
AI Engineer · Build the Right Thing: Product Engineering (Part 2) — Kent C. Dodds, EpicProduct.engineer
Recommendation · MoneyAsk the judge to explain its thinking before it scores — chain of thought measurably improves accuracy because articulating why something is actionable or correct leads to fewer mistakes than picking a label without thinking. The explanation is also very helpful for you as the eval author.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Method · MoneyWiring up a custom evaluator in code is just like wiring up the other evaluators: create a classification evaluator, give it a name (actionability), an LLM, the template, and the choices. Wire up an LLM classification evaluator in code by giving it a name, an LLM, a prompt template, and its choices, then run evaluate on the data frame. The programmatic choices setup mirrors the UI: correct or incorrect, actionable or not actionable, scores one or zero. Programmatic evaluator setup in code is equivalent to configuring the same evaluator in the UI.
Method, not a verified protocol.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Recommendation · MoneyAs your agent matures, turn capability evals into a suite of regression evals that you run all the time to make sure previous behavior has not regressed.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Recommendation · MoneyThe speaker advised writing code evals with no model configuration and no API calls — just Python — for checks like whether the expected stock ticker appears in the report, output length limits, parsable JSON, and forbidden phrases; an LLM judge is overkill for string matching. Code evals can also do complicated things like querying a database or hitting an API to verify stock prices, as long as the grading logic is deterministic.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Recommendation · MoneyGenerate a body of test data — you need more than the one trace you've generated so far to run meaningful evals. Include a bunch of obvious queries that should work, edge cases (e.g., two tickers like Apple and Microsoft for a comparison), and the same tickers multiple times so you can see the non-deterministic output. Different tickers, questions, and complexity levels cover the range of things users will actually ask. In reality, users will ask really weird things you can't predict in advance — another reason evals are important.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Method · MoneyExtract the research context from the traces produced by the first step, attach that output to all spans as the context for the judge, instantiate the faithfulness evaluator in code just like the correctness evaluator, and after attaching context at the column level call evaluate_data_frame with those spans and the instantiated evaluator.
Method, not a verified protocol.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Recommendation · MoneyEval prompts are very sensitive to wording — treating them like code without versioning and testing turns evals from a useful tool into noise. Version your eval prompts and test them on examples where you know the right answer.
AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI
Recommendation · MoneyKarthik recommends storing reusable agent knowledge in a shared memory system rather than empty files, so retrieval costs fewer tokens.
endorsed
AI Engineer · Agent Memory Is Solved. Agent Learning Isn't. — Karthik Ranganathan, Yugabyte
Recommendation · MoneyPienaar's core recommendation: replicate what works in coding by giving the agent critical signal that comes back quickly, so it can ground on it during or at the end of an investigation. In coding, the agent tries a change and gets immediate feedback — it didn't compile, or here's an error message — and that grounding loop is what to reproduce for production, where sync time can be on the order of hours.
AI Engineer · Your AI Agent Is Confidently Wrong About Production — Willem Pienaar, Cleric