Carl 😇 Ngan

125 OpenClaw startups now listed on TrustMRR. $274,397 of verified revenue generated in the last 30 days. If this represents ~1% of the market, @steipete's open source project helped developers make around $30M in February.

76

4 Comments

If your audience can't do something useful with your product in less than 10 mins, you are at a significant disadvantage. No amount of DevRel will fix this.

4

For $10/month, you can use any model out there if you know where to go. The 200 IQ move is not to pay $20 for a single model, but to pay $10 for @Abacus's ChatLLM and get access to every model. You also get a router to engage the best model to solve your problem (like GPT 5, but a router that actually works). You can also share context between all models and use several of them to solve a single problem. IYKYK.

28

7 Comments

𝗧𝗵𝗮𝘁 $𝟳𝟱,𝟬𝟬𝟬 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗶𝗮𝗹 𝗯𝘂𝗱𝗴𝗲𝘁? 𝗙𝗼𝗿𝗴𝗲𝘁 𝗮𝗯𝗼𝘂𝘁 𝗶𝘁. 𝗜 𝗷𝘂𝘀𝘁 𝗽𝗿𝗼𝘃𝗲𝗱 𝘆𝗼𝘂 𝗻𝗲𝗲𝗱 $𝟱𝟬 𝗮𝗻𝗱 𝘀𝗼𝗺𝗲 𝗰𝗿𝗲𝗮𝘁𝗶𝘃𝗶𝘁𝘆. A Hawaiian electrician called me last week. His competitor launched a polished YouTube ad and was winning customers. "I need something like that, but I can't afford agency prices." 𝗤𝘂𝗼𝘁𝗲 𝗵𝗲 𝗿𝗲𝗰𝗲𝗶𝘃𝗲𝗱: 6 weeks, full crew, $50,000-100,000. 𝗪𝗵𝗮𝘁 𝗜 𝗼𝗳𝗳𝗲𝗿𝗲𝗱: "Give me 2 days and $50." 𝟰𝟴 𝗵𝗼𝘂𝗿𝘀 𝗹𝗮𝘁𝗲𝗿... I delivered a commercial featuring a bear mascot in an aloha shirt saving a backyard luau from an electrical emergency. 𝗛𝗶𝘀 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲: "This looks incredibly professional." 𝗥𝗲𝘀𝘂𝗹𝘁𝘀 𝗶𝗻 𝗼𝗻𝗲 𝘄𝗲𝗲𝗸: → 2,000+ views  → 20 new customer calls → $10,000 in booked work → 200x return on investment 𝗧𝗵𝗲 𝗲𝘅𝗮𝗰𝘁 𝘀𝘁𝗮𝗰𝗸 𝗜 𝘂𝘀𝗲𝗱: ChatGPT 4o → Scene generation and scripting OpenAI Sora → High-quality video scenes Kling AI → Image to video scenes ElevenLabs → Voice synthesis and sound effects FAL → Lip-sync technology 𝗛𝗲𝗿𝗲'𝘀 𝘁𝗵𝗲 𝘁𝗵𝗶𝗻𝗴: The $50 covered just the raw tool costs. The real value is knowing how to use them and what to make—that's what agencies charge $75k for. The AI tools make it accessible, but it's the creative strategy that drives results. 𝗦𝗶𝗱𝗲 𝗻𝗼𝘁𝗲: I moved from San Francisco to Honolulu in March and have been working with local clients here. Projects like this are making me love the decision even more. 𝗧𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝗿 𝗽𝗶𝗰𝘁𝘂𝗿𝗲... Small businesses can now create high-quality commercials without traditional production costs. 𝗧𝗵𝗶𝘀 𝗹𝗲𝘃𝗲𝗹𝘀 𝘁𝗵𝗲 𝗽𝗹𝗮𝘆𝗶𝗻𝗴 𝗳𝗶𝗲𝗹𝗱. Local companies can produce content that competes with major brand campaigns. Quality creative work is becoming accessible to everyone. 𝗛𝗮𝘃𝗲 𝘆𝗼𝘂 𝘁𝗿𝗶𝗲𝗱 𝗔𝗜 𝘃𝗶𝗱𝗲𝗼 𝗰𝗿𝗲𝗮𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝘆𝗼𝘂𝗿 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝗼𝗿 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀? Share your experience in the comments 👇 #AIvideo #AImarketing #Hawaii

1

𝗧𝗵𝗲 𝗦𝗮𝗮𝗦 𝗦𝗲𝗹𝗹-𝗢𝗳𝗳 𝗜𝘀𝗻’𝘁 𝗔𝗯𝗼𝘂𝘁 𝗩𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻𝘀. 𝗜𝘁’𝘀 𝗔𝗯𝗼𝘂𝘁 𝗕𝗿𝗼𝗸𝗲𝗻 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗠𝗼𝗱𝗲𝗹𝘀. 𝗣𝗿𝗼𝗽𝗲𝗿𝘁𝘆 𝗠𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝗦𝗵𝗼𝘂𝗹𝗱 𝗣𝗮𝘆 𝗔𝘁𝘁𝗲𝗻𝘁𝗶𝗼𝗻. Over the past few weeks, we’ve watched a sharp sell-off across the SaaS sector. The market message is pretty clear: 𝗔𝗜 𝗶𝘀 𝘂𝗻𝗱𝗲𝗿𝗺𝗶𝗻𝗶𝗻𝗴 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗺𝗼𝗮𝘁𝘀 𝘁𝗵𝗮𝘁 𝘄𝗲𝗿𝗲 𝗯𝘂𝗶𝗹𝘁 𝗼𝗻 𝘀𝗲𝗹𝗹𝗶𝗻𝗴 𝘀𝗲𝗮𝘁𝘀, 𝗹𝗶𝗰𝗲𝗻𝘀𝗲𝘀, 𝗮𝗻𝗱 𝗮𝗰𝘁𝗶𝘃𝗶𝘁𝘆 — 𝗻𝗼𝘁 𝗿𝗲𝘀𝘂𝗹𝘁𝘀. This isn’t new. We’ve seen this movie before. Advertising once sold impressions. Then Google and Facebook rewired the industry. CPM (cost-per-impression) gave way to 𝗖𝗣𝗔 (cost-per-acquisition) and 𝗥𝗢𝗔𝗦 (return-on-ad-spend), and budgets flowed to platforms that could prove impact. Incumbents didn’t fail because ads disappeared — they failed because their pricing models stopped making sense. 𝗣𝗿𝗼𝗽𝗲𝗿𝘁𝘆 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝗶𝘀 𝗰𝗹𝗼𝘀𝗲𝗿 𝘁𝗼 𝘁𝗵𝗮𝘁 𝗺𝗼𝗺𝗲𝗻𝘁 𝘁𝗵𝗮𝗻 𝗺𝗼𝘀𝘁 𝗽𝗲𝗼𝗽𝗹𝗲 𝗿𝗲𝗮𝗹𝗶𝘇𝗲. For decades, PM has operated on a CPM-like model:  • Fees based on % 𝗼𝗳 𝗿𝗲𝗻𝘁 𝗼𝗿 𝘂𝗻𝗶𝘁𝘀 𝗺𝗮𝗻𝗮𝗴𝗲𝗱  • Value measured by activity (rent collected, tickets closed, reports delivered)  • Weak linkage between 𝗳𝗲𝗲𝘀 𝗽𝗮𝗶𝗱 and 𝗮𝘀𝘀𝗲𝘁 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 In plain terms: owners pay for effort, not outcomes. That worked when execution itself was scarce. AI is changing that. Traditional PM providers now face structural — not cyclical — challenges:  • 𝗠𝗶𝘀𝗮𝗹𝗶𝗴𝗻𝗲𝗱 𝗶𝗻𝗰𝗲𝗻𝘁𝗶𝘃𝗲𝘀: fees stay flat whether NOI improves or deteriorates  • 𝗟𝗮𝗯𝗼𝗿-𝗯𝗼𝘂𝗻𝗱 𝗺𝗮𝗿𝗴𝗶𝗻𝘀: scale is limited by headcount, not insight  • Focus on managing process, not 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗶𝗻𝗴 𝗿𝗲𝘀𝘂𝗹𝘁𝘀 AI breaks this equilibrium:  • Leasing becomes a 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗽𝗿𝗼𝗯𝗹𝗲𝗺  • Maintenance becomes 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝘃𝗲, not reactive  • Delinquencies and renewals become 𝗽𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘁𝘆-𝗱𝗿𝗶𝘃𝗲𝗻  • Pricing becomes 𝗰𝗼𝗻𝘁𝗶𝗻𝘂𝗼𝘂𝘀 𝗮𝗻𝗱 𝗺𝗮𝗿𝗸𝗲𝘁-𝗮𝘄𝗮𝗿𝗲 Once outcomes are measurable and optimizable, charging “X% of rent” starts to look like charging per ad impression in 2005. The next generation of winners won’t sell “property management.” They’ll sell performance:  • Fees tied to 𝗡𝗢𝗜 𝗹𝗶𝗳𝘁  • Upside-sharing on rent growth and occupancy  • Accountability for cost avoidance and bad-debt reduction The SaaS sell-off is a warning shot. 𝗣𝗿𝗼𝗽𝗲𝗿𝘁𝘆 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝘄𝗼𝗻’𝘁 𝗱𝗶𝘀𝗮𝗽𝗽𝗲𝗮𝗿. 𝗕𝘂𝘁 𝘁𝗵𝗲 𝗮𝗰𝘁𝗶𝘃𝗶𝘁𝘆-𝗯𝗮𝘀𝗲𝗱 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗺𝗼𝗱𝗲𝗹 𝘄𝗶𝗹𝗹. AI isn’t making PM cheaper. It’s making the old pricing indefensible.

5

Stop overthinking your SaaS auth. Pick one in 30 minutes and move on: If you're building a SaaS today, you face this choice fast. Clerk. Auth0. Supabase Auth. Firebase Auth. NextAuth. Custom auth. I see founders stall here for weeks. I did the same on my last project. I compared pricing tables. Read docs for hours. Debated edge cases that never showed up. Meanwhile, the product sat idle. Here’s the truth. Auth is not your product. Unless you sell security tools, users sign in and move on. So I changed my rule. Default to the fastest stable option. For most SaaS builders today: - Clerk if you want speed and clean UI out of the box - Supabase Auth if you already use Supabase - Firebase Auth if you live in the Google stack - Auth0 for enterprise needs - NextAuth if you run deep in Next.js - Custom auth only if you have a strong reason Why this works: - You launch faster - You avoid maintaining password flows - You reduce security risk - You keep focus on core features Speed beats perfection. Every week spent debating auth is a week without user feedback. Ask yourself: Are you building an auth system or solving a painful problem for users? Pick one today. Ship this week. Let real users test your idea. That is how you win.

3

"Context engineering” is just prompt engineering with added context. The LLM looks at two things: 1. Prompt = Guard rails, Format, Instructions, etc... 2. Context = What documents/code/chat history to include, Summary, etc... How is it different from plain prompt engineering? The queries are heavily augmented here, with chat history summary and other relevant documents. Think of it as RAG + structured prompts. Calling it "engineering" is a stretch. It's too broad and domain-specific to be a science. If you are a software engineer looking to learn more about AI engineering, try this structured course at InterviewReady: https://lnkd.in/gAFcCkRj Cheers! #ContextEngineering #LLMs #AI

55

5 Comments

How to get scammed in software dev space? Pay $63,500 for “special treatment” like this. 👇 An Alabama founder told me his story on a call. 4 months of “development”. At the end he got: 1 Android install file. It crashed on open. No working app. No code he could see. No docs. When he asked for the source code the team said: “Send $5,500 more and we will give you the repo.” That is not a change request. That is a hostage situation. He thought he was buying an app. What he really bought was 4 months of blind trust. No clear milestones. No weekly demos. No written plan for handover. Here is what should happen instead. You see progress every week. You get builds to test. You talk through what was done and what is next. You have in writing that you own the code at the end and when you get it. You might not get everything on day 3. But you should never reach month 4 and see a broken file and nothing else. If a team wants full payment with no clear delivery points no regular demos and only “you will see everything at the end” That is maaaajor red flag. In software you are not just choosing a price. You are choosing who you trust with your product.

6

My app encountered two significant bugs: images were not rendering, and the entire admin page was crashing. Rather than spending hours debugging, I utilized an AI to resolve both issues instantly. This video demonstrates the Lovable AI diagnosing a Supabase private storage bucket problem and fixing broken Supabase database permissions in real-time. This represents the future of software development—transforming hours of frustrating work into seconds of progress. Check out the FULL TUTORIAL where I build this entire app from scratch with the AI! - https://lnkd.in/ebktvbWD #ai #nocode #vibecoding #lovable #coding #programming Lovable

5

Me and my co-founder, Ethan Chen, just finished a 5 hour meeting with one of our customers. In trying to validate Caldrn's idea, we talked to a mortgage broker based in LA, California. We learned some crucial problems that brokers are facing: 1. Time wasted on pushing papers. A lot of a broker's time is spent on organizing client documents and extracting information like D2I and net income for the 1003. 2. Repetitive workflows. Small tasks like going to Redfin to pick up a property's APN and then having to go to the LACounty property website to search up tax reports. 3. A way to effectively communicate with borrowers. Our customer needs their own POS system to keep borrowers up to date with their loan application processes so that there is transparency between both sides. Caldrn.com aims to solve all of these issues with our AI Agentic browser extension for mortgage brokers.

15

1 Comment

The fastest way to signal "we're serious about Flutter" is not your stack. It's your constraints. Senior developers are trying to predict their day-to-day. I see a lot of founders who can't answer these: Release cadence. Performance bar (startup time, jank, app size). Native work expected in the next 6 months. Who owns incidents and hotfixes. If you can't answer those, the role is undefined. Tomorrow I'm posting the 2023–2025 store trend line.

23

1 Comment

I built a full-stack agentic RAG app with Claude Code that can be run locally or in the cloud. I've just released a 2+ hour video walking through the whole thing. You don't need to know how to code to follow along with this; you just need to be technically minded and willing to learn how everything fits together. You'll see my complete development process and how I keep Claude Code on track throughout the build. This is a full-stack RAG app (React/FastAPI/Supabase) with: 🔹 Multi-format document ingestion with Docling 🔹 Hybrid search + re-ranking 🔹 Metadata filtering 🔹 Text-to-SQL 🔹 Web search tools 🔹 Sub-agents with streaming 🔹 Local/cloud model support 🔹 LangSmith observability 🔹 Multi-user auth with RLS Watch the video above and get started using our GitHub Repo (PRD & Prompts), linked in the video description! Link in comments 👇

29

5 Comments