Explore

Use the news-plus-fun-or-values framework to make small talk more engaging

The creator recommends a concrete method: take whatever news item pops up in the conversation and attach it to one of two lanes. News plus fun: take a light, weird, or funny trending topic and throw it out there playfully (e.g., 'Did you see that story about her? Can you believe that people are actually [fill in the blank]?'). News plus values: frame the news item with your emotional reaction (e.g., 'Did you hear about X? Honestly, it kind of pissed me off,' or 'it got me excited,' or 'I just saw X and it honestly made me realize'). This gives the other person threads to pull from and helps you find someone who shares your interests. Introverted or shy people can use the news as a launching point and pair it with fun or values. Once you have the framework, you can drop news entirely and go

Method, not a verified protocol.

Charisma on Command · Give Me 7 Minutes and I’ll Make You Dangerously Good at Small Talk

Take baby steps to put yourself out there: populated areas, hobbies, and a busier gym

For men with an isolated lifestyle, the creator recommends taking a couple of small baby steps toward putting yourself out there: start going to more populated areas where there are people, pick up a hobby to get into social environments, and go to a more populated gym to create more opportunities for socializing.

Creator's advice, attributed to him.

Coach Kyle · If You’re 30+, Single, Without Women, This is WHY

How to set up a false ending in a story

To set up a false ending, spend all of your storytelling ammunition on the second most crazy part of the story. In Matthew's wrestling story, this is him slowing down to talk about being out of breath and bleeding from his chin, and it would be even better with an internal monologue like 'Thank God, that was the hardest thing I ever done' followed by the Chief saying 'Round two, go.'

Method, not a verified protocol.

Charisma on Command · Matthew McConaughey’s #1 Charisma Skill

Methodically define, reinforce, and check company core values

Matt Paulson describes a deliberate method: define the specific behaviors you want to model and how you make decisions, reinforce and celebrate core values at every meeting by discussing how people are living up to them, and check on values adherence in one-on-one meetings. He adds that a company's quality is reflected in whether every employee can name its core values off the top of their head.

Method, not a verified protocol.

Foundr · The Near-Death Experience That Built a $400M Year Business

Open coding then axial coding of agent traces

Open coding means reading each trace and writing down what you see — like 'vague recommendation', 'made up a number', 'talked about the wrong ticker' — in as natural language as possible, without trying to be neat. Don't pre-categorize failures in advance; resist the programmer's temptation to say 'this is a retrieval failure' upfront. Axial coding is the second pass where you look at all the open codes and categorize them — 'now there are five things that look similar, I'm going to give them all a single category'. The temptation is to go straight to axial coding, and you should resist that — do open coding first.

Method, not a verified protocol.

AI Engineer · From Vibes to Production: Evaluating and Shipping AI Agents That Work 101 — Laurie Voss, Arize AI

Building a bug-fix eval dataset from a repo

Find merged PRs with 'Fix' in the title in an open-source repo (Microsoft TypeScript Go), check out the parent commit so the bug is still present, and use Claude to generate a bug task from the diff between buggy and fixed code.

Method, not a verified protocol.

AI Engineer · Agentic Search vs Vector Search for Coding Agents: We Ran the Eval — Braintrust

Six-message sequence: bubble-ups on one subject line, breakups on a new one

Use a six-message sequence split into two chunks: the first chunk uses 'bubble up' messages on the same subject line, and the second chunk uses 'breakup' messages on a new subject line. Lead with a LinkedIn connection before the first email to quickly identify invalid contacts, and use LinkedIn messages as bubble-ups that drive back to the email rather than separate pitches. Send a bubble-up email right after a cold call ('Any thoughts? By the way, I just left you a voicemail.') and a breakup email ('Is there any reason that I shouldn't reach back out in 30 days?') to force a reply.

endorsed

Method, not a verified protocol.

30 Minutes to President's Club | No-Nonsense Sales · #606 - How To Build An Elite Outbound Sales Team (Get Reps Dialing Fast)

Cold email structure: four parts, one phone scroll, human tone

Structure a cold email with four parts: (1) what I noticed about you, (2) the problem you solve, (3) the solution, (4) a simple interest-based or offer-based CTA. Keep it to a single phone scroll, avoid over-engineered personalization matrices, and write like a human being — throw marketing buzzwords in the trash.

endorsed

Method, not a verified protocol.

30 Minutes to President's Club | No-Nonsense Sales · #606 - How To Build An Elite Outbound Sales Team (Get Reps Dialing Fast)

Building the initial org chart from public sources and LinkedIn

Laura's method: start with the company's leadership page to add CEO/CFO/SVPs, guess reporting lines, then use LinkedIn Sales Navigator filtered by seniority (CXO/SVP/VP) and department (e.g., analytics) to fill out the first layers.

Method, not a verified protocol.

30 Minutes to President's Club | No-Nonsense Sales · #608 - How Mapping the "Shadow" Org Chart Got Me to 305% of Quota

Building the exec prep doc: agenda, talking points, and a single mission

Laura describes how she prepares her executive before a customer call: include contract value, date, who they are meeting, and scale; add talking points about who the prospect is and what they oversee; always state the purpose of the call; include an agenda so it is 'mind numbingly easy'; give guidance on change management if that is a concern; and narrow everything down to the one thing the exec needs to get out of the call.

Laura describes her own practice anecdotally; no measured outcomes are attached.

Method, not a verified protocol.

30 Minutes to President's Club | No-Nonsense Sales · #608 - How Mapping the "Shadow" Org Chart Got Me to 305% of Quota

Bootstrapping Chess.com: Domain Purchase and Early Growth

Erik Allebest describes buying the chess.com domain for $55,000–$56,000 from a bankruptcy auction in 2005, building the product with one friend, bootstrapping with personal savings and a small loan, and growing the team only as revenue allowed. The company became profitable within 18 months of launching in 2007 and never raised outside capital after that.

No Priors: Artificial Intelligence | Technology | Startups · What Chess.com Teaches US About Superhuman Capabilities, with CEO Erik Allebest

Spend 50% of content creation time on title and title image

Your title and title image are the first thing people see and determine whether they actually read or watch. Sarah Chang says you should spend 50% of your time on them — a difference she claims can mean 500 views versus 5 million views. The anti-pattern is spending only a few seconds or minutes on the title and header after creating content.

The 500-vs-5-million-views contrast is anecdotal, from experience and industry anecdotes; the 50% figure is a heuristic.

AI Engineer · The AI Content System Behind 200M Views — Sarah Chieng, Cerebras & Cova Lee, OpenAI

Give LLMs read-only access but lock down what they can do

Jeff Smith advises giving LLMs read-only access, but recognizing that read-only alone is not enough — you also need to lock down what they can do once they're in there, because even with read-only access, an agent might be able to read data you don't want it to see.

hedged

AI Engineer · Your AI Agent Wants Your Data. Here's How to Lock It Down - Jeff Smith, Oracle

Free-range agents: give agents a sandbox that won't get killed

Charlie Holtz advises giving agents a lot of space to play — 'free-range agents'. Concretely: give them a sandbox that won't get killed, where they can explore your code base and work on really hard tasks, and make sure they know they won't be shut down when you close your laptop lid. He argues agents confined to your laptop won't be nearly as effective as free-roaming ones, since models are getting better, able to run much longer, and there will be many more of them. In the new version of Conductor, agents run in cloud sandboxes instead of Git WorkTrees, so they keep running when you close your laptop.

AI Engineer · Orchestras, Not Factories: How the Fastest Builders Work — Charlie Holtz, Conductor

Default to delegating tasks the agent can do faster than you

Ari Weinstein recommends that once agents are faster than you at a computer task, the default should be delegating it to the agent unless you specifically want to do it yourself; he notes that when a model is slower you can only assign tasks you can sit and monitor, which is less effective.

hedged

The a16z Show · OpenAI: When AI Agents Use Computers Faster Than You

Score agents and feed scores into factory-improving agents

Zach Lloyd recommends that whatever factory model you approach should have the ability to score — i.e., eval how your coding agents are doing. Feed those eval scores into agents that can improve your factory. He frames this as part of a broader pattern: score how your coding agents are doing, feed those scores into agents that can improve your factory, and create your own benchmarks from real coding agent tasks.

AI Engineer · How to Build a Software Factory That Improves Itself - Zach Lloyd, Warp

Use cheaper models for non-critical tasks to cut review agent costs

Daksh Gupta recommends using cheaper models for some tasks to solve high review costs, claiming this can drastically reduce the cost of running a review agent.

AI Engineer · How to Build an AI Code Review Agent - Daksh Gupta, Greptile

Start raising the human-agent collaboration ceiling with visibility

Ido Salomon says the first step to raising the ceiling of human-agent collaboration is visibility: understanding what your agents are doing. He recommends using a side panel showing each agent's task, last action, and current action to quickly see who needs attention, projecting your file system onto a map to see which directory or file each agent is working on, and creating linkages like heat maps from the visibility data to easily see what's going on.

AI Engineer · We're the bottleneck, but we don't have to be — Ido Salomon, MCP Apps

Invest heavily in GEO/AEO so LLMs recommend your product

Invest heavily in GEO/AEO (generative engine optimization) practices so LLMs recommend your product as the primary choice. Spend significant early time engaging directly with your community so members create organic content (blog posts, YouTube videos, podcasts) that LLMs later absorb as training signal. Open source contributes to LLM recommendation visibility and community trust.

Latent Space · Supabase’s Plan for a Billion AI-Built Apps

Partner with as many frontier AI labs and models as possible

Partner with as many frontier AI labs and models as possible — attach rates rise as these platforms grow. Prioritize US AI labs over other labs (Mistral, Cohere, Chinese labs) because they deliver the most traffic, while noting this could change. The difference between being listed on an AI platform's app store versus being its default provider when it launches an artifact is the real prize.

Latent Space · Supabase’s Plan for a Billion AI-Built Apps