I'm a VC who finds $120 worth of AI subscriptions more productive than an intern. This is my tech stack.

Aug 07, 2026 11:01 AM - 1 hour ago 2

This as-told-to effort is based connected a speech pinch Marc Palet, 26, a Singapore-based task capitalist astatine OMVC. The pursuing has been edited for magnitude and clarity.

I was first exposed to task superior while moving an e-commerce manner business successful assemblage backmost location successful Barcelona. During a study overseas programme successful Hong Kong, I did a ton of networking and came crossed a VC patient investing successful fintech, blockchain, and AI — each the things I liked. I ended up staying pinch them aft graduation and spent 3 years successful Hong Kong earlier moving to Singapore to lead the firm's Asia Pacific investments.

I'm 26 now and person invested successful complete 12 AI companies since I started my VC career. AI isn't conscionable a portion of my portfolio; it's go a important portion of my workflow and really I negociate things arsenic the firm's only worker extracurricular the US.

Even though I've vibe-coded galore of my workflows, AI hasn't replaced immoderate of my package subscriptions. It's really added more, since I salary much than $120 a period crossed Claude, ChatGPT, n8n, and different AI devices for illustration transcription APIs.

But I'd beryllium happy to walk much if it makes maine much productive.

Here's what my tech stack looks for illustration and really it has helped shave countless hours of operational work.

Summarizing laminitis meetings

I started utilizing ChatGPT successful 2023. But the existent displacement came erstwhile I began utilizing coding devices for illustration Claude Code, Cursor, and Replit.

One of the first things I built for activity automatically summarizes my meetings pinch founders.

We grounds our meetings utilizing Fireflies. Instead of manually listening backmost and penning finance notes aft each meeting, I person a system summary covering the company's problem, solution, market, and different accusation we seizure during owed diligence.

My first knowledge base

One recurring problem was that if I weren't astatine my laptop, I'd publication thing absorbing connected LinkedIn aliases X and past hide each astir it.

So creating repositories connected NotebookLM was a turning constituent for me. Every clip I publication thing online that I thought I wanted to remember, I'd transcript and paste it into that repository.

Whenever I wanted to constitute a blog post, I could query it and retrieve each the accusation I needed.

Building a institution brain

The shape I'm successful now started a mates of months ago, aft I publication thing OpenAI founding personnel Andrej Karpathy wrote astir building a knowledge guidelines that automatically ingests accusation and keeps updating itself. I thought that was really interesting, truthful I built thing akin for task superior utilizing Claude Code.

It pulls accusation from Gmail, Slack, my telephone transcription service, newsletters, YouTube podcasts, and anyplace other I get knowledge aliases updates connected our portfolio companies. An AI supplier parses the earthy information, past organizes it into the correct place.

Each portfolio institution has pages for things for illustration fundraising, financials, and commercialized updates. If there's a caller Slack message, the AI figures retired which institution it's astir and updates the applicable page. And it doesn't conscionable transcript and paste caller accusation — it rewrites the page truthful everything stays coherent.

More powerful than an intern

One workflow automatically creates our quarterly constricted partner reports by summarizing everything that's happened crossed our portfolio during the reporting period. Another helps pinch audit activity by gathering accusation from each our portfolio companies.

Those tedious tasks that utilized to return up truthful overmuch of my time. For a VC money arsenic mini arsenic ours, clip is ever a constraint, truthful I'm happy to walk connected AI that tin prevention maine time.

My agents for LinkedIn posts, an important portion of my role, person helped maine summation my output from 1 deep-dive article each 2 months to 1 each 3 weeks. Every article brings successful much woody flow, and because I'm spending little clip connected operational work, I person much clip for in-person meetings.

I realized that AI is overmuch much businesslike than training an intern for a month, only for them to enactment for 3 months. AI besides does things the measurement I want them done. I realized that AI positive maine is much powerful than maine positive an intern.