Recommendation · MoneySending one mass recap email to all six attendees invokes the bystander effect and gets zero responses. Instead, still send the group recap, but also send one-to-one recap emails to each attendee addressing their specific questions — offering the integrations team to the technical questioner or a deeper analytics dashboard demo to the person who asked about it — enabling multithreading and relationship-building that a single mass email never achieves.
30 Minutes to President's Club | No-Nonsense Sales · #625: 18 Deal Cheat Codes to Cut Your Sales Cycle in Half
Recommendation · MoneyEarly in a sales call, ask permission to share what you learned about the prospect's business during prep — nobody will say no, and it lets you present your research.
30 Minutes to President's Club | No-Nonsense Sales · #625: 18 Deal Cheat Codes to Cut Your Sales Cycle in Half
Recommendation · MoneyRichmond Alake recommends getting your semantic layer and ontology layer built out, as they become very important for agent accuracy.
AI Engineer · Everyone Has the Same AI Models. Here's What Sets Agents Apart - Richmond Alake, Oracle
Recommendation · MoneyAri Weinstein recommends that as computer-use agents become superhuman and can no longer be fully observed, enterprises need configurable controls over human-in-the-loop approval and auditability.
hedged
The a16z Show · OpenAI: When AI Agents Use Computers Faster Than You
Recommendation · MoneyMasaba Gupta argues that doing something out of the box and out of the ordinary requires commerce backing it, and points to her own out-of-the-box Nilaya installation as something she could only do because commerce had backed her with funds and creative power.
Figma · Config India 2026: What makes the House of Masaba feel like Masaba
Recommendation · MoneyAs you automate more development, Zach Lloyd says you want to get to a world where you are measurement-first on your own data rather than relying on Twitter opinions or public benchmarks like SWE-bench or TerminalBench. He argues you can't trust public benchmarks as a way of understanding how good a model or harness will be on your own workflows. As an illustration, on Twitter any given day you'll see conflicting claims about which model is best — someone says Pi, someone says Grok — plus benchmarks like SWE-bench or TerminalBench. Instead of just looking on Twitter for model recommendations, he suggests actually testing models on your own system.
hedged
Also discussed by Mario Martone on The a16z Show (unclassified source — not counted until classified). Not counted as independent sources.
AI Engineer · How to Build a Software Factory That Improves Itself - Zach Lloyd, Warp
Recommendation · MoneyThe team endorses leveling up evaluation infrastructure as a prerequisite to scaling recursive self-improvement.
AI Engineer · An Agent That Fixes Its Own Mistakes - Saloni Gandhi & Ziwei Guo, LinkedIn
Recommendation · MoneyHire a critical open-source maintainer of technology you rely on immediately, even against accelerator advice to avoid hiring; and once hired, let them keep building the upstream project rather than pulling them onto your product, so support requests get answered by the maintainer anyway.
Latent Space · Supabase’s Plan for a Billion AI-Built Apps
Recommendation · MoneyKevin Hou advises builders: if you're building a product around an agent, consider what the primitives in your product are and how they might scale with the model's intelligence; with the right primitives, you might be surprised at what the models can do.
hedged
Commercial context: The creator describes this as their own product.
AI Engineer · Get Out of the Model's Way — Kevin Hou, Google Antigravity
Recommendation · MoneyRoland Gavrilescu recommends keeping the information each loop generates — around harnesses, profiles, evals, models, resources, tools, and the environment — portable, versioned, and evolving over time.
AI Engineer · The Loop Is the Product — Roland Gavrilescu, Introspection
Recommendation · MoneyRoland Gavrilescu recommends that the recipe should be something you control, that lives in your company, and is agnostic to the models and providers you use — not tied to any platform.
AI Engineer · The Loop Is the Product — Roland Gavrilescu, Introspection
Method · MoneyCarson describes using a merge queue manager to coordinate the wave of PRs coming in from multiple agents: database migration numbers need to be coordinated, because if all the agents try to grab these things you constantly collide; the merge manager helps collisions go away and is a little more intelligent about how the merge queue is happening; Untangle uses GitHub merge queues, which he says is not rocket science but works pretty well; the manager sessions are just sessions in Devin — nothing special — and he names them so he understands what's happening; set up the roles and then unhobble the model, since having a merge manager say 'hey, make all this better' really works well.
Method, not a verified protocol.
AI Engineer · How One Founder Manages 20 AI Coding Agents at Once - Ryan Carson, Untangle
Recommendation · BodyThe host says a pair of dumbbells and an incline bench are all you need to build big biceps with just two exercises: the incline drag curl and the preacher curl.
ATHLEAN-X™ · The ONLY 2 DB Biceps Exercises You Need (NO, SERIOUSLY!)
Recommendation · MoneyTereza lists being agnostic as one of the three important things in a software factory: stay independent of LLM choices and of how the organization already works. She argues this is best for everyone because new technologies, frontier models, and benchmarks keep appearing, making it hard to catch up and to predict winners.
AI Engineer · Rise of the Software Factory — Tereza Tížková, Factory
Recommendation · MoneyTereza advises that becoming a software factory requires rebuilding the organization from the ground up, rather than inviting a consultancy and dropping something into the middle of an existing organization.
AI Engineer · Rise of the Software Factory — Tereza Tížková, Factory
Recommendation · MoneyTeresa's three important principles for a software factory: be agnostic (independent of LLM choices and of how the organization already works), be autonomous (give agents the right trust, permissions, and governance, and trust them to run for a long time), and always improving (onboard new agents with good codebase understanding, structure, and documentation, and let them improve and share knowledge within the team).
AI Engineer · Rise of the Software Factory — Tereza Tížková, Factory
Recommendation · MoneyInsulate against model availability in different regions or different cloud providers.
AI Engineer · How a Trading Firm Ships New AI Models the Same Day - Noyan Tokgozoglu, Optiver
Recommendation · MoneyWhen you want to improve a prompt, run reflection with an LLM or agent that reads the collected feedback and proposes a better prompt, obtaining the score along with textual feedback that can be very domain-specific, rather than just a reward score.
AI Engineer · Beating RL With Reflection: GEPA and Optimize Anything — Lakshya A. Agrawal, GEPA
Recommendation · MoneyLakshya A. Agrawal advises that if your domain produces expert feedback, compiler error messages, profiler messages, or tool call error messages, you should return that as side information to the optimizer / fitness function.
AI Engineer · Beating RL With Reflection: GEPA and Optimize Anything — Lakshya A. Agrawal, GEPA
Recommendation · MoneyRoman encourages founders to focus on making something that is not possible now possible; users will come to use that thing, and by pulling them to the next impossible frontier a company gains distribution and data advantages.
Lenny's Podcast · How we built Grok Bot in a month | Roman Ugarte (SpaceXAI)