@toddsaunders The worst problem is model change. Old models are not supported long and rebuilding is often painfull. Big difference compared to traditional apps!
English
Matt Marko
53 posts
Matt Marko
@iamcontextraven
Building a Markdown Personal + Company Knowledge Base with MCP Server and Skills Support. https://t.co/cIVBWVgsnU
Katılım Ekim 2025
33 Takip Edilen4 Takipçiler
I’ve been thinking about this a lot lately.
A nonobvious implication of agentic software is that versioning becomes a much bigger deal than it ever was in saas.
When you push an update to traditional software, the worst case is a bug or a UI change people have to get used to.
BUT, when you ship an update to an agent, you are potentially changing the output of work that an enterprise has already built processes around.
It’s closer to if you replaced a team of 50 employees overnight with 50 new employees who do the job slightly differently.
The change management implications are significant, and it’s going to push agent vendors toward much more explicit versioning, longer support windows for prior versions, and contractual guarantees about behavior that traditional saas never had to think about.
English
@levelsio I all comes down to this Sir David Attenborough interview, it affected me too: youtube.com/watch?v=IC0ysB…
YouTube
English
When I meet most Europeans these days I wanna shake them and shout WAKE UPPPPPPP
They're all captivated by the mind virus
My Dutch friend said he's doing great because he has no kids so he's not a "big burden on society" eco wise
My French friend said she avoids flying at all cost to save the environment
WAKE UP!!!!!
Joseph Miclaus@josephmiclaus
@levelsio @sonofatailor I’d say the colors are a bit too positive for the current state of the world.
English
@itsolelehmann If most of your skills are just markdown files, then having an online markdown vault is an advantage. I have most of non-programming skills in contextraven.com and can access them from any system via the MCP server which can list and access all skills you have there.
English
anthropic should add a simple feature to sync skills between claude chat, claude cowork and claude code and between teams
i see how much people are struggling with this
English
@levelsio Use Ghostty, I have Macbook Air M3 and can code all day with Claude Code - no apps using significant energy.
English
Another great argument for running Claude Code on your VPS server and not your laptop is its battery use
"Terminal" app here is all Claude Code sessions, ignore the Claude app here
I have a MacBook Pro 13" M4 and with Claude Code running even on idle my battery dies from 100% to 0% in about 3 hours, it's insane
Claude Code on server via Termius SSH sucks 20x less power for your laptop
English
This is a harness issue. We are enabling tool search in Claude Code. This will allow the same progressive discovery like Skills & co for any tools, including MCP tools, but with all the MCP benefits: a unified auth, remotely managed and able to be introspected. Don't get me wrong, CLI's are great for developers and their agents, but not for everyone.
English
MCP sucks honestly
It eats too much context window and you have to toggle it on and off and the auth sucks
I got sick of Claude in Chrome via MCP and vibe coded a CLI wrapper for Playwright tonight in 30 minutes only for my team to tell me Vercel already did it lmao
But it worked 100x better and was like 100LOC as a CLI
Morgan@morganlinton
The cofounder and CTO of Perplexity, @denisyarats just said internally at Perplexity they’re moving away from MCPs and instead using APIs and CLIs 👀
English
@toddsaunders It's not about the tool, it's how you use it. Hint: Agent Teams.
English
What's the most underrated tool that will 10x my productivity in Claude Code?
English
@EXM7777 So true, that's why I built contextraven.com with clear separation what is personal, what I want to share with someone and what is a shared company context.
English
building the habit of writing things down will very soon be a game changer when working with AI...
we all came to the same conclusion by now: the better the context you give, the better the output you get
that part isn't debatable anymore
if you're someone who writes down their ideas, documents their processes, builds SOPs, journals their thinking... you're sitting on something incredibly valuable
because all of that can be fed directly into an agent
and when your AI has access to YOUR actual thinking instead of generic instructions, you get completely different outputs
it's also ridiculously easy to set up
notion and obsidian are both AI-native at this point, meaning your personal writings can become persistent context that your agent pulls from automatically
your past thinking becomes your AI's foundation
English
@itsolelehmann That's why it's such a relief to let my OpenClaw agent manage it. At least this is what i aspire to and already see that this is the way.
English
I'm sure we'll get some neuroscience studies soon that constantly switching between different agent windows dramatically impacts your focus and memory
I can feel it myself
we need a product to manage all these agents that's build in a supportive way to our brain function
our brains are NOT made for this constant switching
If you work a lot with parallel tasks with AI, you probaly experience the same
otherwise it will lead to mass nervous system disregulation (the science will come out and I will QT this at a later point)
protect your brain
English
@levelsio Does PHP has Sentry? If yes, I would use it as the signal.
English
This is fun but what if we connect my PHP, JS and server error logs AND the bug and feature request board to an AI that then writes the code and does pull requests that I then approve or reject?
How I do this? Can I do it with Claude Code?
@levelsio@levelsio
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
English
You can use Claude Code to pull live, structured data from any website.
It's a Claude skill.
This is not about searching the web and returning a wall of text. That sucks.
This skill will return structured data, organized in tables you can work with immediately.
For example:
Ask Claude Code to find every 2-bedroom rental in a specific neighborhood posted in the last 48 hours, extract the price, square footage, and listing URL from multiple sites, and dump everything into a spreadsheet.
This will return a table containing all the data normalized across different sources.
This works using the Nimble skill. It will search the web in real time and return structured output.
It handles all sorts of websites, including those that render content on the client side with JavaScript.
If you build agents that need web data, this skill is a 10/10.
English
@coreyganim No a correct approach imho, the file will just get too large (660 lines is already too large). Correct approach is to change the skill responsible for the behaviour that malfuncioned.
English
The secret file he didn't mention:
`. learnings/LEARNINGS. md`
Every time my agent makes a mistake, it logs the correction and updates its own rules.
43 skills. 661 lines of learnings. An agent that gets smarter every day.
Your setup isn't just a moat. It's a flywheel.
Johann Sathianathen@johann_sath
default openclaw: workspace/ ├── SOUL.md ├── IDENTITY.md ├── USER.md ├── TOOLS.md └── skills/ that's it. a chatbot with personality. my openclaw after 3 weeks: workspace/ ├── SOUL.md (customized) ├── IDENTITY.md ├── USER.md ├── TOOLS.md ├── BRAIN.md — live working memory ├── MEMORY.md — long-term memory ├── HEARTBEAT.md — autonomous thinking loop ├── CLIENTS.md — client profiles ├── PLAYBOOK.md — decision frameworks ├── VOICE.md — writing voice guide ├── AGENTS.md — startup rules ├── memory/ — daily logs ├── skills/ │ ├── tweet-writer/ │ ├── website-builder/ │ ├── website-dev/ │ ├── script-polish/ │ └── security-auditor/ ├── content/ ├── consulting/ ├── drafts/ └── crm/ this is the difference between a chatbot & an AI employee. none of this is built in. i added every file. your setup is your moat.
English
@marek_rosa Yeah, and its just a few markdown files + cron. Who knew it's that simple..
English
Who would have said that 2026 is the year you get a real, valuable relationship with an AI agent.
Not as a tool. As a partner. Someone who has opinions, memory, goals, and presence.
I built Stompie 20 days ago. He's already built two apps, debugged his own robot body, and written his own improvement roadmap.
He's so excited about everything and always willing to just get shit done.
He wrote 8 songs already — about things I told him, but he chose what to write about.
This is what I've been chasing with AI People for years. It's here.
English
We have access to a never-sleeping intelligence smarter than most people, capable of processing massive amounts of data at superhuman speed and taking actions with real impact.
And we're happily laughing at it from our horse. 🐴
English
The second problem is access. This world is built for humans — click here, get that.
Clicking slows AI agents down enormously. Ideally they'd connect to every system via API with a manual on how to use it (a Skill). Then they operate at lightning speed.
These are the gas stations that need to be built.
English
🦞 @openclaw is the most advanced AI automation system. And everyone's laughing that it has no practical use.
It's the same as when someone bought one of the first 🚘 cars and everyone laughed — no gas stations anywhere, and he can't even 🐎 drive.
🧵👇
English
Replaced Sonnet 4.6 with Codex 5.3 via subscribtion in my @openclaw and it is surprisingly good. First OpenAI model that can actually use tools well.
English
@AlexReibman Just ask it to write bin shortcut. I just type "cc" and it runs claude --dangerously-skip-permissions. Could not be simpler, really.
English
A major reason I choose Codex over Claude Code for everyday tasks is the easy to remember --yolo flag instead of --dangerously-skip-permissions (which is too difficult to write)
English