Thinh Nguyen

Supabase wants to raise funding that values the company at over $5 billion. Supabase's platform, built on PostgreSQL, offers developers tools like authentication, APIs, and real-time subscriptions without locking them into proprietary systems. The company has grown its annual recurring revenue to $70 million and previously closed a $200 million Series D round at a $2 billion valuation. Read more: https://lnkd.in/egMfHD-e 📰 Subscribe to the Daily AI Brief: https://lnkd.in/eq56P3bj #ai #artificialintelligence #ainews #aistartup #aifunding

6

DeepSeek V3.1 just dropped. No press release. No hype cycle. Just showed up on Hugging Face. But look at these numbers: 71.6% on Aider coding benchmark. Better than Claude Opus at 68x lower cost. 66% on SWE-Bench. Competitive with frontier models. 128K context + hybrid thinking. One model, two modes. Here's what matters for production: Your entire codebase fits in one conversation. No more chopping context into fragments that break everything. Two modes, one model. Fast answers for simple stuff. Deep reasoning when you need it. Open source. 685B parameters. No vendor lock-in. Most "AI agents" are just chatbots with function calling duct-taped on. This is different. But here's my worry: DeepSeek has no branded agent framework like Claude Code. The open-source ecosystem will probably mess this up with bad prompts and cheap deployments. You'll see providers serving with lowered precision, pruned experts, poor sampling. Your experience will vary wildly based on where you deploy. Your reality check Week 1: Deploy locally Week 2: Build one autonomous task Document analysis. Code review. Data pipeline. Whatever. But test multiple providers. The benchmarks mean nothing if your deployment is garbage. This is probably as far as they can push V3. But for coding agents? This works. What are you building first?

11

3 Yorum

How I use AI for assisted development at the start of 2026 1️⃣ Discuss with ChatGPT 5.2 the high level design goals and create a spec 2️⃣ Export AGENTS .md file and initial prompt for codex to create a task list 3️⃣ On WSL I run Codex with GPT5.2-codex (High) to create a series of prompts/0x01-task.md files which define scope, spec & criteria for the completion of the project 4️⃣ Review and schedule them in series so Codex executes them one at a time, staged in the background 5️⃣ Deploy, test & create a snag list. Use Codex & Claude to debug, carry out security review and iterate

1

There are two types of vibe coding: "Full vibe coding" and "Semi-vibe coding". Recently had a thought-provoking conversation with Jenny Wong about #VibeCoding. Many engineers are against it, but as a serial CTO from Series A/B startups and someone who has run my own startups, I am NOT against vibe coding at all. 1️⃣ Full vibe coding, with tools like Lovable/Replit/Bolt, is a game-changer for non-technical founders aiming to build MVPs quickly and validate ideas. While critics highlight issues like AI-generated bugs, security holes, and the risk of sky-high cloud bills, these are manageable risks for small-scale, early deployment. The key is to recognize that vibe-coded prototypes aren't meant to scale; when validation is achieved, expect "big bang" refactoring or complete revamp. 2️⃣ Semi-vibe coding, which is already mainstream for experienced teams like ours at Aizzie: we've been leveraging GitHub Copilot since 2021, then Cursor last year, and now Claude Code and Kiro. Here, AI assists us in code generation, but everything remains under expert supervision. This hybrid model enhances productivity while keeping true engineering discipline intact. Embracing both forms of vibe coding can accelerate time-to-market and empower both engineers and non-technical founders, as long as we remain mindful of the trade-offs and transition at the right time.

25

1 Yorum

In just a few years, 🦔 james hawkins took PostHog from an idea hacked together right before YC's W20 deadline to a unicorn powering product analytics for thousands of teams. He joins YC's Brad Flora to talk about surviving six months of "pivot hell," why open-source analytics was the breakthrough, and how PostHog grew from fighting for its first users to launching full product lines—plus what he's learned about momentum, staying close to customers, and using transparency and humor to build a company that stands out. https://lnkd.in/gKSzybvv

173

16 Yorum

I just read the latest piece on Y Combinator’s new startup “wishlist" which you can find here: https://lnkd.in/gxzfycaK Lists like this are always interesting. They give founders a window into where markets may be heading and where investors are paying attention. And for a lot of early builders, that kind of signal can be helpful. But it also got me thinking about the flip side, based on a conversation we had this morning with Aja Beckett and Tyler Blackwell. What happens when founders start with the idea… but haven’t lived the problem? No judgment in that question. Every founder starts somewhere. But startups are hard enough even when you have experienced the pain firsthand. When you haven’t, the learning curve gets steeper: • You’re learning about the customer while building the product • You’re validating the problem while pitching the solution • You’re earning trust while still understanding the workflow That’s a tough place to build from. It’s why “founder–problem fit” matters so much. Many of the most iconic companies backed by Y Combinator came from founders solving problems they knew personally: • Airbnb — needed help paying rent • Stripe — devs frustrated with payments • DoorDash — saw delivery gaps with local merchants That lived experience creates empathy, urgency, and product intuition that’s hard to replicate from the outside. There’s data to back that up. Research from Harvard Business School shows that founders with domain experience are more likely to succeed in that same industry. And according to CB Insights, the #1 reason startups fail is still a lack of real market need. Which says less about founder talent… and more about how hard it is to truly understand a problem from a distance. Startup idea lists are valuable. They spotlight opportunity. But the founders I see build the most durable companies usually start somewhere deeper… A frustration they’ve felt A system they’ve navigated A problem they’ve lived Because when the problem is personal, the commitment tends to be as well. And in startups, that kind of conviction matters.

15

I shipped 2 vibe-coded projects and working on a 3rd. Scaled to 15k+ users in less than a month. - Built on Flutter and a custom backend. - The auth was built with Firebase Auth. Problems: - Context is a problem. Although it is somewhat solved with GitHub Copilot's summarising the chat after context overflow. - Another problem is when you switch projects. So there has to be some manual work in connecting APIs, which is also solvable with the AI generating detailed API docs, and you can copy-paste it in the frontend. - UIs are not consistent, especially complex UIs. So there can be some unnecessary frustration and manual work over there. - The code can easily go out of hand, as the AI doesn't know the entire codebase at once. So the vibe-coder plays a crucial role. Cost and Time: Massive cost and time arbitrage; Considering a developer who knows both frontend and backend, working for a month would charge you at least $1000. With Copilot's subscription and usage-based pricing, that would cost you <$30. The cost can reappear if the project needs to be refactored later. So it's better to be cautious from the start. Time required will be at least 1/3rd, considering you're proactive and on top of everything. Also, if given the correct instructions, it maintains the code quality consistently. With learnings from the past 2 projects, I am doing the current one a little more cautiously. - First, "Plan" what is to be done, and then go into "Agent" mode. - Attach all required files as context for a query. - Create documentation for each module (once they're done building); this helps save context when referring to a module later – like context caching. - Pre-plan the entire project (in terms of schemas, APIs, UI) and only then jump into it. - Not hardcoding API keys inside code and putting everything in .env from the very start. - Create a global instructions.md file with a detailed explanation of how I want the code to be structured and the best practices I want to follow and attaching that to every chat while creating them. - Sometimes, file management goes for a toss. So mentioning the file where I want the code to be written is really helpful More than complete vibe-coding, it was controlled vibe-coding. Treat the AI as an intern who just never complains :)

15

3 Yorum

Terminal Use (YC W26): Vercel for background agents. Terminal Use is hosting infrastructure for background agents that makes harness engineering easy. https://lnkd.in/gv99wN2t

51

2 Yorum

Moved to SF? Don’t start with a startup. Do this onboarding checklist instead: 1. See the sea lions at Pier 39 2. Try Waymo — it feels like the future 3. Walk through the fog Sometimes, the best way to settle in isn’t pitching. (But yeah, eventually… better start building.)

16

4 Yorum

💭 Y Combinator CEO says YC startups launch with 95% of their code written by AI 😮. Harness Field CTO Martin Reynolds brings valuable insight to the conversation with LeadDev around this trend and its impact on software developers. "Having a human in the loop will continue to be vital. The future of coding will involve a combination of human creativity and AI-driven efficiency." Read more on how AI is reshaping engineering roles ⬇️. https://lnkd.in/gnRmrWxS

2

Supabase Launch Week Drives Developer Tooling Momentum Trend: Supabase’s Launch Week introduced storage, edge improvements, and hackathon winners that highlight product growth. Why it matters: Rapid platform improvements reduce friction for building full-stack apps with Postgres at the center. Question: Which platform improvements would most reduce your time to prototype? 🔁 Repost if platform maturity matters for your startup or team 🔔 Follow me for platform updates that cut dev time in half 🌟 Takeaway: Small infra improvements compound into big developer velocity gains

1

We tested our MCP on 21 MCPMark Postgres tasks vs Supabase's MCP. Summary: ✅ 14% higher accuracy (28% for strict Pass⁴ tests) ✅ 1.3x faster ✅ 2.4x fewer tokens Agents build better backends when using AI-native tooling. As models get more capable, that advantage only grows.

22

3 Yorum

Codegen tools are (currently) most valuable/useful for: 1. non-technical product prototyping- vibe coding. 2. non-technical product takeover- founder moding. 3. technical product makeover- vibe shipping. 4. technical product prototyping- vibepreneuring. Non-technical and technical get max leverage for creative problem solving. Idea to product experimentation & iteration. I anticipate an explosion of founders, startups, upside earning collaboration, and unaccredited investors in early-stage ai-first niche to scaled products. May it be known as the coming Slykgeist. And the way of hte Slykigai.

Figma just acquired AI startup Weavy for ~$200M. AI workflows are changing how creative work scales. Here’s why founders should pay attention 👇 Weavy is a Tel Aviv–based AI image & video generation startup. This is Figma’s most important strategic move since Adobe’s attempt to buy Figma was blocked. We’ve entered the AI design consolidation era, where major platforms are becoming AI-first for design and workflow automation: → Canva acquired Affinity (to expand into pro design) → Adobe acquired Rephrase.ai (AI video generation) → Freepik acquired Magnific (AI image upscaling) Because here’s the competitive reality: • Adobe controls the production suite • Canva controls lightweight brand creation • Framer controls design-to-web automation Companies now need to own workflows end-to-end to dominate the market. That means controlling: creation + automation + distribution Because in software, public markets reward leverage. It’s no longer about who has the best design tools but who automates the canvas itself. --- Our team at FE International is more bullish than ever on the rise of AI-first businesses. If you're a founder curious about your company’s true value, get a free valuation. (Link in comments) 👇

34

9 Yorum

I asked YC President Garry a question that changed how I think about building products. Everyone's obsessed with speed in the vibe coding era. Ship fast, iterate faster, move fast and break things. So I asked Garry: "With so many tools making it easy to build, how do you win buyers' minds when they have endless options?" I expected something about going viral or distribution strategies. I'd been telling my friends: "In vibe coding era, there are so many options. You need to build a social following and give people reasons to choose YOU over competitors." But, his answer was simple: "Craftsmanship." I felt ashamed. Here I was, obsessing over personal branding while he talked about fundamentals. I'd been so focused on making people choose "ME" that I forgot to make something worth choosing. He said: "Look at Linear. They really care about fixing every single bug." It reminded me of a concept from my culture: 장인정신 (工匠精神) = the dedication to perfect one's craft This took me back to my childhood. In the early 2000s, my family's jewelry business was in crisis. Chinese factories flooded the market with cheaper, faster production. But my dad had learned craftsmanship in Japan before starting his business. Growing up, I'd watch him obsess over tiny details. When I was impatient to play, I'd ask: "Dad, why are you so obsessed with these tiny details?" My dad would say: "I take pride in my work. When you meet high standards, clients stay loyal for decades." While competitors focused on speed and cost, my dad doubled down on craftsmanship: - Studied thick design books (I was amazed he read every single page) - Spent extra hours perfecting details others ignored - Focused on quality when others chased volume In a few years, competitors closed. He survived. This dedication to craft isn't new. Garry reminded me: "Even Mark, in Facebook's early days, coded 'a Mark Zuckerberg production' onto every blue-and-white page. He was claiming ownership of his craft." Now, anyone can build a quick AI prototype. But building something that lasts still requires: - Deep technical skill - Attention to detail - Time and patience At Plato, Justin and I are committed to this craftsmanship mindset in AI slop. Not just fast. Not just cheap. Built to last. What's your take? Are you building fast... or building to last? P.S. I wrote about Linear's craftsmanship philosophy + practical tips. Comment if you want the breakdown *Thanks Clipboard for hosting such thoughtful conversations!

400

75 Yorum

This Week At Xiaohongshu (Rednote) - 📩 *New Issue Live: 3 Moves That Matter* - AI Power Plays from Global Unicorn Visionaries *Read now* 👉 https://lnkd.in/e4S7visU - Fast-track your AI edge with a crisp, 2-minute brief: *strategy, product, funding.* *High-value intelligence* for serious investors and builders. #ai #rednote #xiaohongshu #Unicorn

4

Supabase vs Firebase — here’s the comparison most devs actually need. Every week I see people ask: “Should I use Supabase or Firebase?” So I put together a video breaking this down from a real production perspective, based on apps I’ve actually shipped using both. What I cover: – Postgres vs NoSQL (and why this matters way more than people think) – Auth, real-time data, and common footguns – Scaling and long-term flexibility – Pricing, lock-in, and when each platform breaks down – What this means specifically for iOS & SwiftUI apps This isn’t a “which is better” video. It’s about which backend fits your app and your constraints. 🎥 Full breakdown here: https://lnkd.in/en-bWXYN Curious what you’re using right now — Supabase, Firebase, or something else?

13

2 Yorum

Nexlayer is part of Jason Calacanis's Founder University Cohort 12. Traditional cloud platforms were built for human developers and DevOps engineers. AI coding agents and vibe coders need something different. We're building it. Nexlayer is agent-native cloud infrastructure — CPU and GPU. Deploy any full-stack app, any language, any backend, any framework, any open-source model from your favorite coding agent (cursor, claude code, gemini etc..) in seconds. No more complexity tax of platform engineering and DevOps. The agentic infrastructure era for AI has started. Time to ship. → nexlayer.com #FounderUniversity #AI #AgenticCloud #AgenticInfrastructure #agentnativecloud #autonomousdeployments

12