Recommendation · MoneyImprove AI design results by leveraging the features that tools offer, such as skills and design-system context. Spend 10 minutes bringing your existing context (design system, components, skills) into the AI tool first, because it will save you hours in the long run. Add your design system and bring in your components as a skill in Magic Patterns. Give the AI guardrails via components so it can output complex experiences like a theme park in one or two shots. Updating a design system is as simple as prompting something like 'no trees, just lollipops and more chocolate rivers'.
Also discussed by CJ Avilla on AI Engineer, Figma, NNgroup, Thomas Watkins on NNgroup (unclassified source — not counted until classified). Not counted as independent sources.
Y Combinator · How To Stand Out In A World Of AI Designs
Recommendation · MoneyMake the primary CTA the number one thing your landing page drives users toward. Make your CTA button large and readable — don't make users squint to find it. On the reviewed site the CTA was unclear, and a top-right button that just says 'Ronin' when the company is called Ego leaves visitors confused about whether it is a product or a model. If you have something important to communicate, it shouldn't matter the angle the user is looking at it — it should be painfully obvious. Make important text like 'free' large and obvious on pricing pages — there's a reason free is giant.
Y Combinator · How To Stand Out In A World Of AI Designs
Recommendation · MoneyThe host recommends proactively communicating data privacy and trust details (what data is accessed, how it's used, retention policies, security) to get users to download and try your AI product. He suggests asking yourself: what is the data we're going to have access to? How are we going to use it? What are our retention policies? Is it secure or not? The products that lean into this and go above and beyond to make users feel super comfortable win. He also says to communicate that the internet doesn't touch the data in any single way — that's what people really cared about.
Y Combinator · How To Stand Out In A World Of AI Designs
Recommendation · MoneyThe host recommends making the CTA obvious and clear on your landing page, labeling the CTA button with a clear description of what it does rather than just the product name (a button that just says 'Ronin' leaves him not knowing what Ronin is), and designing the page so the primary CTA (e.g., 'Download for Mac OS') is the number one thing the user is driven to click. He also suggests checking the page at 100% zoom to verify the CTA is clearly visible and readable, and scrolling through the entire page to verify the main CTA remains prominent and accessible — a main CTA so small it's barely visible at 100% zoom is a design failure.
Y Combinator · How To Stand Out In A World Of AI Designs
Recommendation · MoneyTanmay Sah recommends extending traditional ML testing (assumptions, input data, behavior; held-out data with metrics like accuracy and AUROC) to the whole deployment system: test what happens when a tool fails, whether retrieval brings back the right information, whether human review catches mistakes, whether the agent stays within its permission boundary, and whether the system can recover. The logic of testing stays the same for agentic systems; only the boundary of what is tested gets bigger.
guest's recommendation
AI Engineer · 1,368 AI Failures and What They Teach About Evals — Vishal Srivastava & Tanmay Sah, New York Life
Recommendation · MoneyEvan Conrad argues that to make a compute market work, you have to operate the cluster like a hotel — run the whole thing — otherwise you can't do physical settlement. This means building out the entire operating capacity of a GPU cloud, and operating both the market and the clusters at the same time.
asserted
AI Engineer · Why AI Startups Will Disappear Without Compute Markets — Evan Conrad, San Francisco Compute
Recommendation · MoneyEvan Conrad argues that to solve compute market problems, you need to underwrite the value of the GPU cluster rather than underwriting the credit risk of the customer — otherwise the compute market will not work.
asserted
AI Engineer · Why AI Startups Will Disappear Without Compute Markets — Evan Conrad, San Francisco Compute
Recommendation · MoneyStart discovery calls by hunting for business problems, not product or situational issues — use broad questions to surface the deeper root cause behind the stated need. When a prospect is fixated on a surface-level need (e.g., 'we need sequence writing'), ask how they got there: 'Why are you struggling with X? Who wrote it? What are your open reads?' Prospects who state a specific need will almost always answer 'what prompted that need?' with a story — the fastest track to the bottom line. Sellers often avoid asking about the actual reason the prospect showed up out of fear, but asking directly is faster and better.
30 Minutes to President's Club | No-Nonsense Sales · #619: Active Listening to Loosen up Your Discovery Calls and First Dates | Sam McKenna | 30MPC Hall of Fame
Recommendation · MoneyActive listening is the number one thing B2B buyers want from sellers — and sellers are terrible at it. Ask a question, wait for the answer, then ask a follow-up active-listening question. When someone shares something ('it was busy'), follow up on their content: 'What made it so busy? Did you expect it to be so busy? Is there something about this time of year that makes it bananas?' Don't respond to a personal share by pivoting to your own agenda (e.g., 'Cool. What do you want for dinner?') — that's selfish focus, not active listening. Practice at home with kids, spouse, partner, neighbors, or parents if you're not good at it.
30 Minutes to President's Club | No-Nonsense Sales · #619: Active Listening to Loosen up Your Discovery Calls and First Dates | Sam McKenna | 30MPC Hall of Fame
Recommendation · MoneyThe source recommends setting up an attribution model and UTM links to track how much each creator is earning, making it so creators can see what's going on with their earnings and performance, and showing creators the top-performing content from the brand so they can make better decisions.
endorsed
Also discussed by Eugene Teo on Garage Strength (unclassified source — not counted until classified). Not counted as independent sources.
Foundr · The NEW Creator Playbook that Took Me from $80K Debt to $250K/Day | Trybe
Method · MoneyThe speaker's team watches every submitted video start to finish and denies any video they don't genuinely think has all the parts of a great ad. Approved ads must have a hook, body, call to action, and unique value propositions, plus no echoing, good lighting, and good editing — if any of that is missing, the video is denied.
endorsed
Method, not a verified protocol.
Foundr · The NEW Creator Playbook that Took Me from $80K Debt to $250K/Day | Trybe
Recommendation · MoneyAt the end of the process, set up automated posting through an API-connected tool like Webber rather than posting manually. Tell ChatGPT to take a product's docs and set up API posting, since many consumer products have APIs. Example prompt: 'Here's a folder of all these posts. Post them one every Tuesday. Schedule it for all these different platforms. Figure out the best time for each platform. And write some captions that I approve ahead of time.' One cut-down can post every Tuesday for six weeks across five channels, with captions sent for approval first, and separate skills maintained for TikTok versus Instagram versus YouTube Shorts.
Presented as the speaker's own workflow and endorsement.
AI Engineer · The AI Content System Behind 200M Views — Sarah Chieng, Cerebras & Cova Lee, OpenAI
Recommendation · MoneyIt is very obvious when captions are AI-generated, so the speaker recommends studying your voice and brand, differentiating voices (personal vs. company) and formats (long-form interview, technical demo, short conversation), using a swipe file of bookmarked tweets as reference, and pasting the Wikipedia 'Signs of AI Writing' article into the tool with the instruction to do nothing mentioned in it.
Presented as the speaker's endorsed workflow.
AI Engineer · The AI Content System Behind 200M Views — Sarah Chieng, Cerebras & Cova Lee, OpenAI
Recommendation · MoneyRun a skill weekly with the script: 'Review my video creation process. Show me where I spend the most time and suggest repetitive tasks I could turn into tools or automate.' The rationale: busy people don't have time to look back and find inefficiencies, but building better systems lets you focus on the creative work that gives you joy.
Presented as the speaker's recommended practice.
AI Engineer · The AI Content System Behind 200M Views — Sarah Chieng, Cerebras & Cova Lee, OpenAI
Method · MoneyUse an AI agent to schedule your cut-down posts across platforms — for example, one cut-down every Tuesday for six weeks across five channels. Tell your posting agent to research the best posting time for each platform itself, have it write captions for your approval ahead of time, and instruct it to publish every Wednesday and send captions for approval first. You can post content through Webber by giving it a folder of posts and instructions to schedule them, and set up posting APIs by telling ChatGPT 'Take this docs and do it for me and set it up.'
Method, not a verified protocol.
AI Engineer · The AI Content System Behind 200M Views — Sarah Chieng, Cerebras & Cova Lee, OpenAI
Recommendation · MoneyDesigners need to write the layer of specs that tells us what the product is supposed to be, while technical folks cover implementation details, security, and data models.
NNgroup · Specs Are The New Source Code (feat. Thomas Watkins, NN/G Guest Instructor)
Recommendation · MoneyMax Kanat-Alexander recommends determining where software sits on the disposability scale using four factors: consequences, cost of replacement, dependencies, and cost of maintenance. If all four factors are low, it's disposable software; if all four are high, it's durable software. He advises erring on the side of software being durable, and erring on all four factors being higher than you expect them to be.
AI Engineer · Does Code Quality Still Matter? It Depends on This — Max Kanat-Alexander, Capital One
Recommendation · MoneyMorgan Suo recommends measuring what the finished job cost and how long it took, including retrieval, tool calls, and retries, rather than judging a model by raw speed alone.
AI Engineer · Is a Faster Model Actually Faster? Test the Whole Job — Morgan Suo, MiniMax
Method · MoneyBoot the three repositories (containing everything FleetKit needs, including FleetKit source) into a Claude Code session with a plugin, invoke it, and give it a mission. The agent adopts or creates a Slack channel for the mission where all war-room members and decision makers communicate, and can ping individual people based on the roster. Everything from the agent goes into a human-in-the-loop queue (e.g., GitHub issues) that humans can comment on and approve and other agents can add context to. Only when everything looks tight from an approval standpoint does the fleet manager dispatch builders to do the rest.
Method, not a verified protocol.
AI Engineer · One Day, $1,000 in Tokens: An AI War Room for Legacy Code — Ford Prior, GEICO
Recommendation · MoneyHave agents log their decisions and what they're doing in an opinionated, standardized way, all checked into Git — so there's a durable record of why decisions were made and the context around conversations. Don't rely on code alone as the memory of a codebase; you need docs to know why you made a decision. Keep all context in the repo, never in someone's head — this gives agents and tools superpowers by having all the context in one place, enabling more agent-driven work.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess