Luke Dz (@LukeDz8) header
Luke Dz (@LukeDz8) profile picture
San Francisco
Joined August 2021
Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)Photo by Luke Dz (@LukeDz8)
My parents were both teachers - so I really can't help teaching all that I've learned about startups. "Speed" is the subject of Masterclass #2. Let me know what you like or what can be improved! @nfx
🚨 It's official: NFX Masterclass is back! ALL episodes are now live. 🚨 Season 2 is all about speed. In fundraising. In building. In marketing. We've been developing this content for years. It's everything you need in 10 new episodes from @JamesCurrier. Free, as always. Watch it now: nfx.com/masterclass/
11
2
30
3,625
scholar-inbox.com daily feed of research papers, upvote/downvote to refine recommendations
1
15
We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case). I think we'll see more of this pattern inside companies. The frontier models are amazing generalists. But increasingly you'll see specialist models out perform on more narrow tasks, either working on teams, or by themselves. We're making it easy to fine tune models internally for our employees. We even have a skill for it in our internal harness that makes it easy to spin these up and start training them: "/fine-tune-and-serve-llm I want to fine-tune Qwen for payments risk review. help me get started" I imagine we'll have some backend services that could be replaced entirely by this new architecture.
614
788
547
9,478
1,611,845
😂😂🔥☝️
2
54
Luke Dz retweeted
Hey everyone, we made some updates based on feedback from you. We’ll keep shipping, so I’ll post more updates soon - connect with your friends to get trip ideas - AI search in your own words (this is just a baby step towards agentic AI, which is a UI coming soon) - get laundry done, rent baby gear, and more
136
45
24
978
226,127
we just launched You on @dateknown, so proud of the team ❤️
3
4
105
3,911
Luke Dz retweeted
Yesterday I was in Austin, TX to announce the Airbnb Housing Accelerator, a $250 million commitment to help get more homes built. Airbnb started because of a housing problem. My roommate and I couldn’t afford our San Francisco rent, so we rented out airbeds in our apartment. Today, housing is unaffordable for millions of people around the world. If you step back and ask why housing has become so expensive, the biggest problem is pretty simple: we don’t build enough of it. America is short more than five million homes, more than all the housing in New York City and Los Angeles combined. We build fewer homes today than we did in 1950, even though the population has doubled. Airbnb has been part of the housing debate, and we understand why communities have concerns. Dedicated Airbnb listings are a small share of the housing supply—about 0.5% here in Austin and 0.3% across the U.S. Restricting short-term rentals is an understandable response, but the evidence shows it doesn’t address the underlying problem. New York City effectively banned Airbnb, and rents have risen about 30% since then. Austin took a different approach and passed reforms that made it easier to build more housing. Rents there are down 18%. Every city is different, but Austin is a clear success story. We believe building more homes is essential to making housing more affordable. So we asked ourselves: what can Airbnb can do to help? We found a few places where we can. Today, around 750,000 homes in the U.S. are ready to break ground, but are stuck because they can’t get the final piece of financing. That’s where we can help. We’re doing four things: 1. Putting $250 million behind housing projects that unlock $5 billion in development. 2. Supporting local housing advocates working to modernize zoning, speed up permitting, simplify building codes and reduce unnecessary costs. 3. Creating a $5 million Housing Innovation Prize for technologies that can make homes faster and cheaper to build. 4. Launching the Airbnb City Index to make it easier to see which cities are building housing, becoming more affordable and adopting policies that work. Our first project is in Austin. We’re investing $6.4 million to help build 201 affordable apartments in St. John, a neighborhood where the land has sat empty for nearly 20 years. The community fought to turn this lot into affordable housing and won. But construction stalled for nearly five years because the developer was a few million dollars short. Airbnb came and played a small part to unlock this project. None of these Austin apartments will ever be on Airbnb. Short-term rentals aren’t permitted. We’re investing anyway, because the goal is simple: help build more housing. We’re starting in the U.S. because it’s the market we know best. I want to prove this works here before taking it further. But housing affordability is a global problem, and I hope we can eventually take what we learn here and apply it globally. This is new for us. We’ll get some things wrong. And $250 million won’t close a five-million-home gap on its own. I know there will be skepticism about this, and that's okay. Airbnb only works if communities trust us and feel good about our presence in them. We can’t ask people for that trust. We have to earn it. The best way to do that is to help solve problems in the communities where we operate. That’s what the Housing Accelerator is about. Brian
210
132
77
1,624
242,362
We are seeking new contracts for all of San Francisco’s homelessness services, prioritizing accountability and results. This effort will shift all contracts to an outcomes-driven model and incorporate performance metrics. We are also seeking feedback from the community and providers throughout this process. Our administration is fundamentally transforming how the city tackles homelessness and addiction. We are not just getting people off the streets—we are helping them get on the path to recovery and stability.  But that work can only succeed with a system that is built to deliver results, with accountability when the outcomes aren't good enough. We are not going to continue spending $1 billion of taxpayer money on a failing system—we're fixing the system so we can deliver the safe and clean streets every San Franciscan deserves.
370
268
144
4,643
609,201
🔥🔥🔥🔥🔥🔥🔥@TSpirit_Dota2 🔥🔥🔥🔥🔥🔥🔥🔥🔥 GOAT
Congrats @Team__Spirit!!! Collapse is the best dota player in the world. // Коллапс — лучший игрок в доту в мире.
1
2
119
🔥🔥
The teamspeak makes this fight even more crazier 😱
15
Luke Dz retweeted
🏆 TEAM SPIRIT ARE YOUR TI 2026 CHAMPIONS! 🏆
14
230
28
1,054
23,141
Best ? to get to know someone quickly is “what’s your biggest pet peeve”
45
🇨🇴💥
1
34
🔥🔥🔥 it’s time to learn
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.
132
In honor of 50 years of Apple, we're sharing - for the first time ever - Don Valentine's original 1977 memo for Sequoia's investment into Apple Computer. #Apple50
185
879
344
6,004
2,274,401
Use @claudeai to learn SOTA: - Get podcast transcripts - Ask Claude to read a transcript & all external links referenced - Have it take notes & use them to build 3 example projects, explaining tradeoffs - One transcript per session - Build new project w/ cumulative learnings
61
Replying to @nikitabier
@nikitabier add recommendations/“for you” per list (rec’d tweets/accounts)? Too cost-prohibitive to build this as an X app using the dev API. Cost should be max $10/person/mo
1
1
1
46
@elonmusk plz add App Store for X apps
24