Mikki Kobvel

@Jacobsklug Let's hang, we don't do lovable do AI-Native Engineering but happy to get some work together

I'm in Barcelona for the month. Locked in. If you're around, have a private rooftop we can work at.

@dexhorthy great talk! love it

👇 a curated superthread of resources to get the most out of coding agents, advanced context engineering, research/plan/implement, and more

@garrytan Best time since JS kicked off things forward

If you are in startups or tech It is literally cook or be cooked

I remember these old days when you had fatigue from drilling to much single bug until 2 am, so you become unproductive but you afraid to lose the context. Well, now similar fatigue but you have 10 tabs of agents working and suddenly you cross-chat with them.

@karpathy @Goss_Gowtham Would be great to learn from practical video

@Goss_Gowtham I was just thinking the same thing

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

Vibe coding is more addictive than any video game ever made (if you know what you want to build).

@sflorimm @PeterSweden7 Eleven labs ?

so far Europe is leading with: 1. Mistral 2. ASML 3. bottle cap 4. cheese 5. Wind Turbines that @PeterSweden7 mentioned

USA has ChatGPT USA has Grok USA has Claude USA has Gemini USA has Llama USA has Copilot China has DeepSeek China has Qwen China has Ernie China has GLM China has Kimi China has MiniMax Europe has?

This is big Change your old dummy email :D

Product Engineering is the most important concept since DevOps era in software development space.

I feel like writing your own posts with typos and mistakes feels more valuable than: clarity, noise, moat slop of AI. I would pay premium to have zero ai people’s opinions

@milanm_ @levelsio @TermiusHQ You can't use those in terminal, that's the issue

@kobvel @levelsio @TermiusHQ As I mentioned above, it feels expensive. I would pay maybe 5. Which is kind if stupid, because I use dictating a lot. But then again, you get it for free in Claude, ChatGPT

Got the 🍋 Neo to try it as a dumb client with only @TermiusHQ installed to SSH and solely Claude Code on VPS No local environment anymore It's a new era 😍 (Oh and a 💅 pink Neo for gf)

@milanm_ @levelsio @TermiusHQ Wispr flow not bad

@levelsio @TermiusHQ What do you use for dictating? Claude's voice mode is good, but not working with VPS setup.

@levelsio @renatonitta @TermiusHQ double that, just make it live on every prompt

@renatonitta @TermiusHQ Why do you need a local dev environment in the first place?

I got so tired of LLM hallucinating about what is today workout. Built an AI agent for Telegram / Web - showing my prep workout plan. #AI #HYROX #BuildInPublic

@noahkagan If you can't sell it and it boosts you already, what's the point of shouting about it?

50% of the people here are LYING about how they use AI and the other 50% aren't talking about it.

Marked2 - set it default for your system (Mac) to preview markdowns. Helps to open .md after agent generates Blazing fast preview of implementation plans

@zackbshapiro It is actually true also for Software Agencies, AI-enabled specialist with his critical thinking is the value not the "Ai can do it" statement

@matheusml This is gold, re-reading the same articles, re-watching the same videos

I switched everything (podcasts, YouTube, audiobooks) to 1x speed. New rule: if it's not worth consuming at 1x, it's not worth consuming at all. terriblesoftware.org/2026/01/08/lif…

@noahkagan Have been using since the inception, Code was a clear winner from the beginning

Shocked how quickly I moved over from ChatGPT to Claude for all my AI usage. Anyone else?

@SahilBloom @ycombinator @Heypocket @AkshayNarisetti @gabrieldymowski how is it different from @PLAUDAI ?

@ycombinator @Heypocket @AkshayNarisetti @gabrieldymowski I’ve been loving this thing. I got it when I was trying to break my phone addiction and have been using it nonstop as a way to log ideas and thinking without needing to touch my device. Thinking and brainstorming sessions with it are great. (Not an investor, though want to be!)

Pocket (@heypocket) is your notetaker for real world meetings. In the last 5 months, the team has delivered over 30k units with a $27M annualized run rate, growing 50% month over month. Congrats on the launch, @AkshayNarisetti and @gabrieldymowski! ycombinator.com/launches/PaX-p…