mattboustred

@thenanyu connecting Claude to Posthog, Supabase, Slack and using it to understand your users better was a massive recent unlock for me. been using CC and OpenClaw heaps but the ease of following hunches in the data through claude has surprised me with how enjoyable and powerful it is

4. PMs and product marketers live out of an LLM. Use MCP to read context and write data PMs and marketers should be doing upwards of 80-100% of their work through a chat interface. This can take all sorts of forms: Notion AI, Claude, Slack, etc. Start with Claude, it has the best MCP implementation so you can communicate with your systems of record and you can upgrade to Cowork once comfortable.

There's an incredibly wide gulf between software orgs on the leading edge of AI adoption compared to the median company. Here's my POV on what every software team that’s doesn’t know where to start SHOULD be doing today— with basically zero friction or FDE

Honestly they all go so hard

Apple’s rotating background where every hour it shows a new photo has made me want to use my phone again. such an enjoyable experience seeing different memories pop up. sometimes the most important feature decisions are about the experience, not the capabilities

@levelsio Next unlock is treating yourself like an LLM… protect your context window, clear it deliberately (meditation), fight the urge to dopamine hunt by switching tasks/consuming unnecessarily The analogy holds disturbingly well

This week I decided to just permanently switch to running Claude Code on the server mostly on bypass permissions mode: c() { IS_SANDBOX=1 claude --dangerously-skip-permissions "$@"; } And for the first time in my life I think I've actually managed to outrun my todo list What happened is I simply blasted through my to do list of features I had to build and bugs I had to fix I've never shipped so fast and Claude Code almost made no mistakes, and when it did it they were tiny that weren't fatal (important because I'm mostly working on the server in production now) Before I was always known to ship fast (also because I always work alone) but while I shipped new things would always build up on my features/bug board (my users can submit them there) But this is the first week where I've been fast enough to outrun them The board is actually empty! As other people have written on here the real bottleneck is becoming myself and my creativity, not how fast I can ship. Because I think I ship faster now than I can come up with new ideas, or maybe my brain will adjust to this new speed (probably) Also I feel another limit is becoming my own mental context window, as in how many things, features, bugs, projects, I can keep in my mind in parallel while building on all of them. It's a lot and I haven't reached that limit yet but I feel I might be close I also noticed that you start going really fast the more you let it just go loose, before I was slow because I didn't trust it and I was scared it would destroy my code, now I just let it go. As @karpathy wrote, things feel like they've changed a lot around December last year when models became good enough to really code with and I feel the same When I see other friends code with Claude Code I often notice they're slow because they still check everything, which is good of course, but I feel the better way would be to create some tests and just let it run freely and see if it can pass those For me the tests are mostly just me checking out if the new feature on the site works or not, and in 99% cases it just does, and then I ask it to improve it further Because I run Claude Code on the server in production, I don't have to wait for deployment anymore (although that took only 3 seconds anyway before, that still adds up), now it's wait for it to be done coding, I refresh the site and I test it, that feedback loop is how I work and it's made me WAY faster Anyway here's what I did this week and the majority of these things were requested by people on the bug board, I'd say this is about 10x my normal output: 📸 Photo AI - Built new image viewer and mobile image viewer - Added batch remix, multi-photo import, filtering by model in gallery - Security overhaul: phased out insecure ?hash= login, migrated to session tokens - Fixed Google login loop, multi-model selection, talking scripts - Added custom audio upload for talking videos - Created dynamic model selector from server endpoint 🏡 Interior AI - Revived [ Add furniture ] feature (started 6 months ago, image models now good enough) - Added custom style upload for redesigns - Built own Gaussian Splat viewer for 3D - Made /remove_bg endpoint for furniture backgrounds - Migrated 3D walkthrough to new World Labs API - Added .skp file support, paint color masking, empty room button 🎒 Nomads - Launched weekly AI-generated newsletter from chat - Built profile edit modal, moved profile editing from /settings to profile page - Added TikTok/YouTube links, status bar, server-side API tracking - Added hundreds of new profile tags and traits - Fixed timezone filters, broken links, user avatars 🗺️ Hoodmaps - Revived write mode (before was only read for last few years because db was rekt) - Built heatmap mode using sentiment-scored tags (50K+ tags) - Fixed root cause: tags not entering DB due to wrong PRAGMA (should be WAL) - Added good/bad area detection with admin grid controls - Set up Claude Code Telegram bot for live changes - Enabled CF cache, fixed health check, fixed Brussels 📕 MAKE book - Built auto ePub/PDF generator cron worker - Added dynamic generation with personal customer watermarks - Added image compression for file size 💾 Pieter .com - Added Wikipedia text-only reader for Kindle - Exploring Windows 3.11 emulator using v86 (to replace Em-DOSBox) - Added product recommendations on homepage - Installed Wall Street Raider (1986) 👩‍💻 Remote OK - Installed Chatbase AI customer support bot - Added "report not remote" link on job posts 🏨 Hotelist (3 todos) - Fixed hotel URLs and city range bugs - Added iron amenity

two years of daily chat taught me less about AI than two months with claude code two months with claude code taught me less than 2 weeks with OpenClaw the mechanism gets more visible as the tool gets more powerful. that's where the learning is

four years as CEO, I never wrote a line of production code. two months ago I shipped our entire Prompt Cowboy rebuild - frontend, backend - mostly alone. the gap between idea and execution used to require a person. now it requires persistence.

a lot of normie (mainstream) talk of an AI bubble, but today felt like living in the future. > in bali > spend an hour planning with cursor > let it loose to build > go for a swim > repeat if that isn't the promised land I don't know what is.

@marcas_daniel curious where you found your chef? I’m in Bali atm and was looking into this to be able to lock in with no distractions, but gave up after we weren’t able to find someone decent and realised all the restaurants over here are pretty affordable anyway

I invested in a chef for my villa in Bali. Here's the ROI: Last month, I did the math on where my time was going. Turned out I was spending 2-3 hours daily on food-related decisions and tasks: - What to eat - Grocery runs - Cooking (which I hate) - Cleaning up That's 60-90 hours a month on something that drains my energy and isn't my strength. So I hired a chef. Here's what changed: - My mornings start with coffee already made. Lunch appears at noon. Dinner at 5pm. All healthy, all handled. - The mental bandwidth I used to spend on "what's for lunch?" is now redirected to client work, strategy, and building my business. The unexpected benefit? - My energy levels improved. When you're eating well consistently (not just grabbing whatever's convenient), you feel the difference. The lesson: - Calculate what your time is actually worth. Then look at what's draining it. For me, hiring a chef to gain 60-90 productive hours monthly was an obvious trade. That time went straight back into revenue-generating work. Not everyone needs a chef. But everyone has tasks that drain them. What's yours? And what would it cost to get it off your plate?

@mattpocockuk this is dope - will have to try it out

Instead of markdown files, I've been getting my AI to save plans in GitHub issues. Claude Code is AWESOME at using the GitHub CLI. Means you don't have endless THROWAWAY_PLAN.md files spamming the repo.

has anyone nailed the figma > code workflow? having so many issues pulling figma designs out for our webapp even when everything is properly tokenised and designed... so much easier to just vibe it out straight into code with cursor

@denny_zhou also clearly written by AI…

we live in a parallel universe lol

@gearside i didn't realise how big Charlie Kirk was until this happened, but the shocking thing to me is how many people are celebrating his death, not realising how morally bankrupt they are being

charlie kirk was someone i often disagreed with. but i respected him. he showed up and literally set up an open mic to opinions he didn’t agree with. he was a totem of free speech; an open mic in human form. we argue, we go home safe. that’s the deal. 
and last night, the deal broke. it didn’t just feel wrong. it felt like something bigger cracking. zoom out. in the book of human history, freedom is not the default setting. violence is. open speech, scientific method, equal protection under the law - these are weird little hacks we’ve patched onto the system. they only keep running if someone keeps maintaining them. and we’re not maintaining them. speech is getting softer. safety is breaking down. standards of living are plateauing, if not falling. a ceo murdered. a political activist gunned down. teenagers killed with machetes in melbourne. arrests in the uk for contentious comments. this isn’t “the world going crazy.” this is entropy. systems decay if they’re not kept up. truth gets massaged. institutions bloat. moral compasses glitch. power shifts from the people who earn trust… to the ones who can control speech. it’s not just america and the uk; australia isn’t immune. our freedoms feel normal. they’re not. they’re exceptionally fragile. this operating system - western society - is worth protecting. but to do so, we need to appreciate what we’re protecting. so ask yourself this: what do you actually value about western society? is it free speech? the sanctity of truth? presumption of innocence? independent media? elected leadership? the right to say something unpopular? whatever your list is - write it down. name it. then pick one thing. and stand up for it. defend it in your life. not in the comments section - out loud. in public. when it matters. because if we don’t stand up for it, who will?

@minmintymin @Coventuresvc dream team

Drumroll, please… After three years as Australia’s first solo female General Partner, @Coventuresvc is expanding from one to two partners. I’m incredibly proud to welcome Alexey Mitko to the team as Partner. He brings the systems, scaffolding, and calm certainty that help founders get through the early chaos. Alexey’s spent fifteen years shaping high-growth companies. He's seen Canva at twenty people, scaled Koala, and co-founded Eucalyptus into a company of hundreds. As an angel, he's invested in dozens of early-stage companies. Anyone who's met him knows: he's the kind of investor who writes the check and sits with you until the messy parts are solved. If you’re thinking about building or have an ambitious pre-seed idea, we'd love to chat.

@nikunj This is a green flag. Almost impossible to know what tf you’re talking about in AI atm without getting your hands dirty

Founders often ask me - “Aren’t you a VC? Why do you keep shipping side projects”.. For me, the best way to learn about a space IS to build it - especially in software. In the last 12 months, I have finetuned models, run large inference on clusters to try open source TTS models, have my own embeddings, built many websites that show me the limits of vibe coding, built and ran my own evals, built plenty of AI agents using frameworks etc. All done on nights and weekends, btw - days are for founder meetings. But, now when I talk with a founder now, I can show them what I have built and dig deep on the challenges that I could foresee in their space. It also sparks new ideas on where a moat could lie (distribution or technical) - something market research or customer references often don’t help with. And I have personally never been afraid of looking stupid - so you’ll continue to see me ship ideas on my X!

@HenryYin_ @richmondalake @MongoDB @agihouse_org memory is also super important when it comes to codifying “know-how” in an enterprise. E.g. how to do tasks that humans have learned by experience but that isn’t written down anywhere

The next big leap of agent adoption will come from memory and personalization. We dug into this on the latest episode with @richmondalake from @MongoDB at @agihouse_org. The “wow, it just gets me” experience comes when an agent knows what to remember and how to use it. Here are 4 types of agent-centric memory and how they unlock better UX: 1. Episodic (conversational history) - Example: A support bot that recalls you already tried solution X last week. - How it helps: Avoids repeating itself, feels continuous. - When to use: Assistants, customer support, coaching agents. 2. Procedural / Workflow (steps and outcomes) - Example: An ops agent that remembers last week’s deployment failed at step 3. - How it helps: Doesn’t retry the same broken step, learns which path worked. - When to use: Multi-step agents for automation, IT, or enterprise workflows. 3. Semantic / Knowledge (facts and references) - Example: A policy agent that pulls the exact discount rule from docs. - How it helps: Reduces hallucination, improves trust. - When to use: Any knowledge-grounded agent (QA, internal tools, compliance). 4. Persona / Preference (identity and style) - Example: A personal assistant that always answers in concise bullets because you prefer it. - How it helps: Feels personal, consistent, on-brand. - When to use: Personalization, brand alignment, long-term relationships. We also discussed the marriage between MongoDB and @tengyuma's @VoyageAI . The plan: bring embedding models natively into MongoDB, then move toward a more holistic RAG solution. The vision is to free developers from plumbing (choosing embedding models, wiring rerankers, manual chunking) so they can focus on building products and UX. Takeaway: Bigger context windows won’t solve memory. The leap forward comes from choosing the right memory type and structuring it thoughtfully. That’s how agents go from functional to magical. Watch full episode here: youtube.com/watch?v=664Gpi…

Things cooking

@maxmarchione I go for molecularly distilled fish oil. Once you find a high quality brand it’s hard to go back to having fishy burps an hour later

the models contain capabilities that literally nobody knows about yet. experimentation and curiosity are your two biggest weapons when building with LLMs

we're in the "command line" era of ai interfaces. those who master the raw interaction now will have massive advantages when more intuitive interfaces emerge.

the things nobody's talking about with ai... the gap between average and elite prompt engineers is growing exponentially. what used to be a 20% difference in output quality is now 200-300%. the best prompters are getting results that look like magic to everyone else.