David Gasquez
LightNews LightNews
banner
davidgasquez.com
David Gasquez
@davidgasquez.com
5.4K followers 1.4K following 1.2K posts
Data @ Protocol Labs.

Open Data, Open Source, Open Protocols.

Walks taker. Progressive Metal enjoyer.

davidgasquez.com
Posts Replies Media Videos
davidgasquez.com
Used @ap.brid.gy to request Nik to bridge his Mastodon posts into Bluesky! Smooth experience and happy to have the posts here. 😊

bsky.app/profile/ludi...
September 1, 2026 at 3:36 PM
Reposted by David Gasquez
softwaredoug.bsky.social
This pattern of asking a cheap LLM to generate hypothetical / fake classifications that we then later resolve to real ones is one of the dominant way I'm doing query understanding these days

softwaredoug.com/blog/2026/08...
The hallucinating classifier pattern
A useful pattern for LLM classification at scale is to let it hallucinate plausible, fake entities. Then resolve to real ones.
softwaredoug.com
August 10, 2026 at 8:52 PM
davidgasquez.com
This quote is even more relevant these days with all the "Company Brain" talk going around.

Everyone is talking about Retrieval Augmented Generation, but most companies don't actually have any internal documentation worth retrieving. Fix. Your. Shit.
davidgasquez.com David Gasquez @davidgasquez.com · 22d
Every now and then, I think about this awesome post and imagine how the author must feel with all the increased AI insanity going on right now on every corner of most companies.

ludic.mataroa.blog/blog/i-will-...
I Will Fucking Piledrive You If You Mention AI Again — Ludicity
ludic.mataroa.blog
August 26, 2026 at 2:17 PM
Reposted by David Gasquez
hannes.muehleisen.org
Some big news: @ducklabs.com is joining AWS, with the move expected to complete by early September.
Our team will stay together in Amsterdam, with more resources to take DuckDB, DuckLake and Quack further - while the open-source Duck Stack remains free under MIT license

ducklabs.com/news/2026/08...
DuckLabs to Join AWS, Projects to Remain Open Source
DuckLabs provides services around the DuckDB in-process OLAP data management system directly from its main developers.
ducklabs.com
August 26, 2026 at 1:03 PM
davidgasquez.com
Every now and then, I think about this awesome post and imagine how the author must feel with all the increased AI insanity going on right now on every corner of most companies.

ludic.mataroa.blog/blog/i-will-...
I Will Fucking Piledrive You If You Mention AI Again — Ludicity
ludic.mataroa.blog
August 26, 2026 at 8:53 AM
Reposted by David Gasquez
hazelweakly.me
The two hardest problems in Computer Science are

1. Human communication
2. Getting people in tech to believe that human communication is important
December 24, 2025 at 12:42 PM
davidgasquez.com
Wrote about why context engineering is a data problem.

Organizations need to own the process that turns raw sources into useful artifacts for their agents.

Much of this is work data teams already know how to do!

davidgasquez.com/context-engi...
Companies want your data and context because [your context is their moat](https://x.com/samzliu/status/2080210797465379147). Products that own your context can force you to use their agent... and, I've never seen a good hosted agent! Organizations need to own their context, and the way to do so effectively already exists and is well understood. ## Context as Data Infrastructure Have you noticed the [shape of the diagrams](https://storage.googleapis.com/gweb-cloudblog-publish/images/WorkspaceIntelligence.max-2200x2200.png) that every company is adding to their "intelligence" products? I see that and it seems like I'm looking at a Fivetran or dbt product page in 2018. That is because they're the same thing! Most organizations will need to build and maintain a model-agnostic knowledge base ([aka ontology, company brain, ...](https://x.com/DBredvick/status/2078150905078206789)) for the same reasons they maintain a data warehouse. A curated and normalized layer helps both humans and agents make sense of all the structure that outlives any particular model, harness, tool, or product. Context engineering is [that same work](https://x.com/JoshARosen/status/2084693306722705629): extracting, filtering, curating, modeling, and publishing artifacts to help the organization make better decisions. For now, it seems there isn't a great set of tools (or [modern context stack](https://x.com/davidgasquez/status/2082895547677954466)) built for this purpose. Something like Fivetran and dbt optimized for the new sources (Slack, Drive) and transformations (transcription, summarization, text extraction from a slide deck, ...).
August 7, 2026 at 7:01 PM
Reposted by David Gasquez
ronentk.me
Great post! There is a clear parallel to how the @atproto.science ecosystem can disrupt the extractive science publishing industry & bring on millions of researchers with a new cooperative research stack.

The system is ripe for disruption - see also @edhagen.net's post in the thread linked below >
August 7, 2026 at 1:44 AM
davidgasquez.com
One of the best things I did recently was to archive all @gordon.bsky.social's Substack posts into my personal digital library. I index, embed and query it with github.com/tobi/qmd.

When my agents now $consult-library they stumble with many great Gordonisms, making them better for my taste!
Screenshot of a YAML configuration in a dark-themed code editor. It defines a squishy-computer source with a local archive path, Markdown glob pattern, commented Substack export command, and context describing Gordon Brander’s essays on tools for thought, cybernetics, complex systems, decentralized protocols, network society, and AI agents.
August 1, 2026 at 9:58 AM
Reposted by David Gasquez
ssp.sh
I still hope AT Proto will win in the social media race. It has such great ideas and architecture behind it, and it's just there, for everyone to query.

Great talk with @danabra.mov, as always.
ATPROTO isn’t just for Bluesky weiners
YouTube video by Syntax
youtu.be
July 31, 2026 at 2:11 PM
davidgasquez.com
@atproto.com (open data networks in general) allows users to build all kind of personal features or views apps themselves may never have an incentive to build.

An app doesn’t have to decide that a certain use case is large/profitable enough to support/build it.
July 31, 2026 at 11:47 AM
davidgasquez.com
Wanted to take a peek at @semble.so and didn't knew where to start. I built this as a static explorer of all the links in there.

sites.wisp.place/davidgasquez...

Love AT Protocol for letting me build the UI I want! 😍
Interactive “Semantic Field Map” for the Semble Link Atlas, showing 31,367 links grouped into 28 color-coded topic clusters. Thousands of colored dots form a central map labeled with themes such as Experimental Web Projects, AI Industry Criticism, Language Model Engineering, Books and Literature, and Open Source Infrastructure; a searchable topic legend with link counts appears on the right.
July 31, 2026 at 10:47 AM
davidgasquez.com
"All happy families are alike; each unhappy family is unhappy in its own way"
vickiboykis.com Vicki @vickiboykis.com · Jul 16
There are like only three design patterns you need to use 80% to make stuff good and fast and elegant

Make it a lookup (hashmap)
Split it up into smaller pieces (sharding)
Combine it into one large piece (map reduce)
July 16, 2026 at 2:05 PM
Reposted by David Gasquez
davidgasquez.com
Write more ad-hoc ephemeral workbenches!

They help you learn and a higher bandwith way to communicate with your agent.

davidgasquez.com/ephemeral-ag...
When I work with agents, my first instinct is usually to “just prompt” through the task. That is fine for many coding-adjacent tasks, but not everything I do can be easily codified, and I bet the same thing happens to you! I’ve explored how to make agents do subjective work better, and I think there is another underrated approach to using agents for subjective tasks, or tasks where the goals are fuzzy and you’d like to keep the final call: build local and ephemeral workbenches. The idea is to have agents produce small pages/apps for the task at hand. For me, the clearest recent example is a grant review workbench I built. Instead of asking a model “which application is better?”, I got it to build me a workbench to help me better understand the task. In the workbench, I could see and sort every application, add random tags or notes as I read through them, visualize them on a 2D map to spot duplicates or core applications, etc. The useful part was the handoff. I could click a button to copy the state and ask my agent to do things like complete the labeling from my manual labels, recluster, add a new field to applications, suggest due diligence questions, etc. This gave both of us a high-bandwidth way to share state and made the feedback loop faster than chat. Chat alone is a bad interface for this kind of work because too much state stays implicit, or even hidden from you. With the workbench, I could inspect both the applications and the changes the agent wanted to make.
July 15, 2026 at 3:17 PM
davidgasquez.com
Wonderful post by @summerscope.bsky.social! Hat tip to @mariozechner.at.

The Human-in-the-Loop is Tired.

pydantic.dev/articles/the...
The Human-in-the-Loop is Tired
On reward functions, dopamine, and what it actually feels like when the code starts writing itself
pydantic.dev
July 16, 2026 at 10:37 AM
davidgasquez.com
Write more ad-hoc ephemeral workbenches!

They help you learn and a higher bandwith way to communicate with your agent.

davidgasquez.com/ephemeral-ag...
When I work with agents, my first instinct is usually to “just prompt” through the task. That is fine for many coding-adjacent tasks, but not everything I do can be easily codified, and I bet the same thing happens to you! I’ve explored how to make agents do subjective work better, and I think there is another underrated approach to using agents for subjective tasks, or tasks where the goals are fuzzy and you’d like to keep the final call: build local and ephemeral workbenches. The idea is to have agents produce small pages/apps for the task at hand. For me, the clearest recent example is a grant review workbench I built. Instead of asking a model “which application is better?”, I got it to build me a workbench to help me better understand the task. In the workbench, I could see and sort every application, add random tags or notes as I read through them, visualize them on a 2D map to spot duplicates or core applications, etc. The useful part was the handoff. I could click a button to copy the state and ask my agent to do things like complete the labeling from my manual labels, recluster, add a new field to applications, suggest due diligence questions, etc. This gave both of us a high-bandwidth way to share state and made the feedback loop faster than chat. Chat alone is a bad interface for this kind of work because too much state stays implicit, or even hidden from you. With the workbench, I could inspect both the applications and the changes the agent wanted to make.
July 15, 2026 at 3:17 PM
davidgasquez.com
Complexity is the mind killer.

github.com/PhilipK/arti...
Complexity is the mind killer TL;DR: When faced with a choice, always pick the simplest thing that solves your immediate problem. Then make sure you can change your mind later. By the simplest thing I mean the solution that is easiest to reason about. When we always pick the simplest solution we avoid the analysis paralysis of finding out which solution might be the best in all the future scenarios we can imagine. We simply pick the simplest thing, and move on, knowing that it is easy to change our mind later. After we implement the simple solution, we have a much better understanding of our problem and we can ask "is my immediate problem solved?" If not; repeat. Complexity is the mind killer. Our mind gets overloaded when we try to change a complex system, and this causes us to make mistakes. Systems must be changeable, it is what keeps them from becoming legacy and it keeps us agile and adaptable. A lot of complexity gets added in the name of changeability. We create microservice systems so we can change our programming language and service implementation, but we add the complexity of network, service discovery, message queues, container orchestration and so on. It also adds a resource overhead that forces us to scale horizontally earlier, which is complex. Could we simply have hidden the implementation behind an interface or a function? Will we actually change our programming language any time soon? Does it actually solve our immediate problem?
July 13, 2026 at 8:48 AM
Reposted by David Gasquez
duckdb.org
Check out the Awesome DuckDB repo – curated by @davidgasquez.com – to explore the growing list of handy tools and cool projects built around #DuckDB, DuckLake and more. 😎 🦆

Notice a great tool that’s missing? Submit a PR to get it added.

Start exploring: davidgasquez.github.io/awesome-duck...
July 9, 2026 at 6:51 PM
davidgasquez.com
The primitive is the product. Great read on the power of extensibility and thinking in terms of abstractions and capabilities.

www.amplifypartners.com/blog-posts/t...
The primitive is the product | Amplify Partners
Why every software company is now a dev tools company.
www.amplifypartners.com
July 1, 2026 at 3:01 PM
davidgasquez.com
Came across this old XKCD. Not only works for sports commentaty these days! 😅

xkcd.com/904
Two stick figures talk. Caption: “A weighted random number generator just produced a new batch of numbers.” One says, “Let’s use them to build narratives!” Bottom caption: “All sports commentary.”
July 1, 2026 at 9:45 AM
davidgasquez.com
Integration complete! My posts are now on atproto too thanks to @standard.site.

davidgasquez.com
davidgasquez.com
David Gasquez's blog
davidgasquez.com
June 30, 2026 at 3:41 PM
Reposted by David Gasquez
danabra.mov
i want to spread the word about atproto. if you’re running a podcast, i’d be happy to hop in to talk about it. i understand it decently well and can explain it without much jargon. it’s one of the most interesting developments in the last 30 years of web, and i can defend that. please dm to set up!
June 30, 2026 at 2:48 AM
davidgasquez.com
Good example! UMAP rules.

bsky.app/profile/gran...
davidgasquez.com David Gasquez @davidgasquez.com · Jun 17
Agents are everywhere now but is worth remembering that there are still so many cool things we can do with good ol matrix factorization and embeddings. No LLMs required.
davidgasquez.com David Gasquez @davidgasquez.com · Jun 17
How community notes reduce viral misinformation. Great talk/conversation about community notes internals and impact.

www.ted.com/talks/keith_...
June 26, 2026 at 2:51 PM
Reposted by David Gasquez
adrimaqueda.com
🌡️ Cada vez que llega una ola de calor me desespera lo difícil que es saber los récords de temperatura. Así que he montado un mapa que rastrea los récords de temperatura de todas las estaciones de AEMET en España y muestra los días que han pasado desde el último récord
June 24, 2026 at 6:22 PM
Reposted by David Gasquez
eugenevinitsky.bsky.social
Proposed norm: you're allowed to respond to an unedited LLM document with an unedited LLM document that is twice as long
June 24, 2026 at 9:20 PM