Gokul Rajaram

Particula has a unique superpower. They have the ability to combine robotics and software expertise to create delightful, addictive consumer hardware/software games. And to do it over and over again. First with GoDice, then with GoCube, then with GoChess and then again with GoBoard. Pretty incredible. I've gotten hours of joy (and learning) playing with their connected games. And when I love something, I try to support it. Consequently, I'm very excited to support Particula in their mission to bring learning and joy to the world through connected gameplay.

Particula has a unique superpower. They have the ability to combine robotics and software expertise to create delightful, addictive consumer hardware/software games. And to do it over and over again. First with GoDice, then with GoCube, then with GoChess and then again with GoBoard. Pretty incredible. I've gotten hours of joy (and learning) playing with their connected games. And when I love something, I try to support it. Consequently, I'm very excited to support Particula in their mission to bring learning and joy to the world through connected gameplay.

tldr Get a venture investor who puts themselves on the line for you. As a founder, you sacrifice your life, your identity, your peace of mind, and more, for a decade or longer. Venture investors, by definition, have a portfolio. As a consequence, they are not as single mindedly committed to your startup as you are. However, that doesn’t mean they can’t be invested. And by investment, I don’t mean capital. That’s cheap and easy, especially today. I mean: - their brand - their time - their networks - their reputation You want someone on your side who’s willing to lay it all on the line for you. To speak for you in rooms you’re not in. To pick up the phone at 11pm at night or 3am in the morning. Capital is the cheapest thing an investor can give you. Ask for the expensive stuff. Before signing a term sheet, please reference check the person (not just the firm) with other startups to see what, besides capital, they’ve put on the line.

A Markdown File is an Employee Garry Tan, President & CEO of Y Combinator, interviewed by Anish Acharya (The a16z Show) Video: https://lnkd.in/gW3VVCTC Transcript: https://lnkd.in/gfx8VABm Tan argues founders can encode every repeatable process, turn organizational memory into software, and operate with context once trapped inside management layers. The result: tiny teams, hundreds of skills, and companies that improve whenever an agent errs. 1. Trust the territory, not the map. Tan left web programming because mobile looked hot, then rejected an early Palantir role to chase a Microsoft promotion. Both choices ignored firsthand knowledge for prestige. 2. Earnestness is courage. When the crowd says your market is dead, maturity is not agreement. It is trusting direct experience strongly enough to keep building. 3. Founders need a truth network. YC’s deepest value is a community where someone can admit the best customer left or a co-founder lost hope. Most startup events produce two people claiming they are “killing it.” 4. Become 400× yourself. Vibe coding and agents let one founder carry the output of a former department. Ambition should rise with leverage; copying the SaaS successes of the last cycle aims at the wrong game. 5. SaaS is a wedge, not a moat. A per-seat tool may open the door, but durable value must compound through proprietary data, network effects, or a learning loop. 6. Build trivial things. Develop AI intuition by using every model on low-stakes projects. The artifact barely matters; it is a chassis for learning what the technology can now do. 7. Agency and taste are muscles. Cheap production makes judgment scarcer. Build both by choosing, shipping, seeing what fails, and choosing better next time—not by waiting to feel qualified. 8. A markdown file is an employee. Perform a process well once, capture the trace as a skill, and schedule it. When the agent fails later, fix the file once and keep the improvement forever. 9. Automate the bottleneck. Tan’s engineering loop automated product work, testing, and QA in sequence. Each time human attention became the constraint, he encoded that step too. 10. Startups must reorganize. Agents can track dependencies, surface conflicts, and retain more context than middle management. Incumbents may struggle to rebuild around this model; a startup can, and Tan says every startup must.

Looking for someone exceptional in Azerbaijan. One of my top portfolio companies is in advanced discussions with the Azerbaijani government around a potentially significant engagement. They’re looking for a highly entrepreneurial, high-agency person to lead the relationship on the ground and, over time, potentially build and run the company’s operations in the country. The ideal person is unusually resourceful, commercially strong, trusted in senior circles, and comfortable operating with a lot of autonomy. This could be someone already in Azerbaijan, or someone from Azerbaijan who has built experience abroad and is excited by the idea of moving back for the right opportunity. This is not a conventional country-manager search. We’re looking for someone who can create momentum from a standing start and become the company’s senior presence in the market. If someone exceptional comes to mind, I’d be grateful for an introduction. DM or email me. gokulr at gmail

Design Only the Durable Ian Silber, Head of Product Design at OpenAI, interviewed by Lenny Rachitsky (Lenny’s Podcast) Video: https://lnkd.in/gPPEECHK Transcript: https://lnkd.in/gyk_jBUW Designers are unhappy because the job’s boundaries are dissolving as tools accelerate. Silber’s answer: use AI everywhere, preserve human judgment where novelty and empathy matter, and spend craft only on decisions likely to endure. 1. Nobody is behind. AI design practice is so early that starting today puts you ahead of most people. What failed last month may work now; coding agents went from frustrating to transformative during one paternity leave. 2. Throw it into an agent. Start an idea in Codex or ChatGPT Work before opening the traditional design stack. A rough artifact produces better questions than another abstract discussion. 3. Roles blur; functions endure. Designers, PMs, and engineers will overlap more, but products still need product judgment, design craft, and engineering rigor. At startups, hire generalists; at scale, keep the hats explicit. 4. AI is already a good designer. It can generate competent flows and visuals. The human edge is finding unmet needs, watching people struggle, inventing interactions without training data, and bringing a point of view. 5. Hire curiosity, not AI tenure. OpenAI does not require an AI background. It looks for curiosity, rapid learning, prototyping, systems thinking, and the judgment to make powerful models useful to ordinary people. 6. Systems beat screens. When engineers ship thousands of changes, designers cannot polish every surface. Build reusable primitives, and extend an existing component before creating another. 7. Design the durable. ChatGPT may test 100 versions and discard 99 for foundational interactions. Other ideas go from concept to production in four hours. Spend craft where technology will not invalidate the decision next month. 8. Build in public at the edge. Codex can expose ideas early to expert users, generate real feedback, and harden before reaching a billion people. Not every experiment belongs in the default experience. 9. Design for the spectrum. One user asks for dinner ideas; another automates a farm. Invite the novice without slowing the expert, and expose advanced capability without forcing everyone to understand models or modes. 10. Fix forward. IGTV flopped; Reels worked. The lesson is not to avoid misses but to ship, listen, recover quickly, and keep learning. Outcomes matter more than process theater.

The Moat is Better, not More Riley Brown, AI creator and founder of Chorus, interviewed by Peter Yang (Behind the Craft) Video: https://lnkd.in/gJR34XWg Transcript: https://lnkd.in/gPQW5MbF Riley Brown has built an audience of roughly 1.7 million by teaching AI. His core idea: agents should improve creative quality, not merely multiply output. He builds reusable skills from real work, combines them into personalized workflows, and keeps human curiosity and taste in control. 1. Better, not more. Most people use AI to batch content. Riley uses it to put more research, iteration, and craft into each piece. Volume without judgment becomes soulless. 2. Build skills from real work. He does not design a grand automation system upfront. When AI completes a useful task he expects to repeat, he says, “Turn this into a skill.” 3. Compose the toolbox. One skill researches videos, another finds assets, and another makes motion graphics. The advantage comes from connecting specialists around a workflow, not building one giant agent. 4. Substance before packaging. Riley chooses a topic he genuinely wants to explore and records the core content first. Only then does he engineer the hook, title, and thumbnail. 5. Script the opening. He designs the first 30 seconds and lets the rest stay conversational. His BRENS checklist asks whether the opening feels Big, Relatable, Easy, New, and Safe. 6. Research as a multiplier. His YouTube-research skill retrieves transcripts, studies successful formats, and turns them into hook outlines. AI lets him put more total research behind one video. 7. Prompt with references. For thumbnails, Riley places successful examples and his own images on a visual board, then asks AI to combine them. Showing the target often beats describing it. 8. Evaluate outcomes. For low-risk creative work, he runs a skill, judges the artifact, explains what failed, has the agent revise it, and retests in a clean chat. The durable discipline is evaluation, not prompt folklore. 9. Keep humans at the edges. AI provides starting points; Riley and his team apply taste through editing, thumbnail iteration, A/B tests, and final judgment. 10. Stay in the arena. Riley’s consulting and product work feed his content with firsthand learning. Distribution without practice eventually becomes commentary about other people’s work.

Manik Gupta (Manik Gupta) is one of the most experienced execs and leaders in tech. Product Leader at Google, CPO at Uber, CVP at Microsoft. He has seen every kind of title and org structure out there. So it's super awesome that he is eliminating titles completely (including for BUSINESS roles) at his NewCo, and every person will have the title Member of Technical Staff. I love this because the MTS title sets the requirement that every person at the company should be skilled at using AI to do their job. Thank you Manik for leading the way and hope other companies follow suit!