michaeld123

If you need more data to help drive features: https://www.linguabase.org

that's not how I interpreted the whole article.

  Location: Takoma Park, MD, USA
  Remote: Yes
  Willing to relocate: No
  Technologies: All LLMs, Python, JavaScript, TypeScript, Swift 
  Portfolio: https://www.michaeldouma.com
  Email: m5@inotherwords-studio.com

I help make performant systems easy to use, both the first time and the 1000th. Last year I built Linguabase, a semantic graph of 400K terms and 63M relationships from 130M LLM inferences. I work across the stack, from systems architecture to highly usable design, and I've led cross-functional teams of 9 to 40. Currently an AI Engineer at the U.S. Treasury. Previously ran a word-game studio funded by a $295K NSF SBIR. Earlier work includes SpicyNodes (40M users), ProstateCalculator (neural-net health AI, 1.5M patients), and time.gov. I judge games for IGF, MAGFest, CODiE, and GEE.

I'm good at communicating clearly: my interactive article on multi-word expressions in English trended on Hacker News in February.

Looking for an applied AI role at a smaller company where I can help you strategize, build, and deploy. I'd shine on developer tools, learning platforms, knowledge products, creative AI, thinky games, or anything that needs to be more accessible to users.

  Latest: https://www.linguabase.org, https://www.inotherwords.app
  LinkedIn: https://www.linkedin.com/in/michaeldouma/

  Location: Takoma Park, MD, USA
  Remote: Yes
  Willing to relocate: No
  Technologies: Python, LLM pipeline orchestration (130M+ API calls across Claude/GPT/Grok), JavaScript
  Portfolio: https://www.michaeldouma.com
  Email: m5@inotherwords-studio.com

My interactive article on multi-word expressions in English trended #1 on Hacker News in March. I help make systems easy to use, both the first time and the 1000th.

Currently at the U.S. Treasury as an AI specialist, but I want to get back to a smaller team. I've led cross-functional teams of 9 to 40. Last year, I launched Linguabase, a semantic graph of 400K terms and 63M relationships from 130M LLM inferences. Before that, I ran a word-game studio funded by a $295K NSF SBIR. Past work includes SpicyNodes (40M users), ProstateCalculator (neural-net health AI, 1.5M patients), and time.gov. I judge games for IGF, MAGFest, CODiE, and GEE.

Looking for an applied AI role at a smaller company where I can build. Developer tools, learning platforms, knowledge products, creative AI, brain health. Anywhere the core problem is some version of 'this is powerful but nobody can figure it out.'

  Latest: https://www.inotherwords.app
  LinkedIn: https://www.linkedin.com/in/michaeldouma/

Both your points are solid. I think I'm pretty liberal about opposites, I can see opposites to major political and cultural figures too. You are right that the mahjong presentation loses rigor if there's not enough downside from choosing the wrong sequence. The mobile mahjong games often have a small fill-tray so the player needs to really focus on the sequence.

the older generation of vectors, like word2vec, compress to one sense of a word, but even if we ignore polysemy, opposites have a lot in common. So for any antonym pair, they would not be minus-1, they would be actually pretty close.

The big challenge is when we go beyond antonyms with clear scales like heat, speed, size. To me "as of now" is recent, and therefore opposite to "old". I would like a word other than "opposite" or "antonym" or "contrast" that captures a wider range.

Historical look back at old articles on this popular word game. February 15, 2022 - bigthink.com Long before the Wordle mania, there was the crossword puzzle craze. And newspapers around the world condemned them as an “invasive weed” that caused mental illnesses and even murder.

  Location: Takoma Park, MD, USA
  Remote: Yes
  Willing to relocate: No
  Technologies: Python, Swift, JavaScript, semantic systems, LLM orchestration
  Portfolio: https://www.michaeldouma.com
  Email: m5@inotherwords-studio.com

I had an article trending on HN last week about "words with spaces" in English. I can help you make your complicated system be more usable for your consumers.

I have a track record that spans from govt services (I created time.gov in 1999), visualizations (SpicyNodes, 40M users), AI-based health info before AI was a thing (ProstateCalculator, 1.5M patients), and association-based word games (two games, plus a game-centric game studio, NSF funded w/ $295K SBIR). I also judge games for several awards (IGF, MagFest, CODiE, GEE, Serious Games).

How can I help you? Maybe your project needs an overall product vision, or incremental improvements that help first-time and long-time users know and (enjoy!) using your product.

I would love to work with your team in a product/design leadership role, at a smallish-company. I can work with developer tools, learning platforms, knowledge products, brain health, creative AI.

  Latest: https://www.inotherwords.app
  LinkedIn: https://www.linkedin.com/in/michaeldouma/

Claude is AI and can make mistakes. Please double-check responses.

And I attempted to add your 'svart hull' note.

I just adapted your comments into the paragraph starting with "German". Hopefully accurately.

My bad. there's a little sidebar about it, but I put it lower after the chart because there wasn't room. You might still not find my logic on the 15% satisfying, but it's there.

It's not so much the prompt, as the volume. This overall project has involved >100M LLM inferences, spread across 1.9M headwords. the building block is "what words or short terms are related to X?", but scaled out. Plus a lot of filtering. So it's mostly a reflection of English, and also a reflection of what ChatGPT and Claude report back as a significant collocation.

Ha — you're probably right that it would have been less controversial. But I kept it precisely because it's arguable. Added a parenthetical acknowledging the HN debate and framing it as on-the-fence by design

Added a note: "'I love you' isn't opaque, but it's tight enough to put on a tile." The familiar end of the spectrum picks up collocations that are transparent but loaded — I'm not claiming they're words in the traditional sense, but they're useful vocabulary for word games, which is where I'm coming from.

You were right — it is now. Thanks

Yeah — added a note below the slider that the obscure end is noise (LLM artifacts, jargon fragments, Wiktionary debris) and that where to draw the line is up to the reader. It was always intended to show the full gradient including where it breaks down, but that wasn't stated

Added Japanese nettō alongside the Slavic examples. Thanks for the specifics

This was a great detail — added Russian kipyatok and Polish wrzątok to the article as evidence that "boiling water" carries enough conceptual weight that other languages crystallized it into a single word

Added Dutch "creditcard" as another example. Thanks

Added a German/Norwegian section — but vidarh corrected me below: German doesn't 'remove the space,' the compound never had one. Adding a space changes the meaning or breaks the grammar. The article now reflects that.

Great example — I added svart hull to the article as an illustration of a language that writes it as two words but still puts it in the dictionary because the meaning isn't obvious from the parts. That's exactly the instinct English lacks.

You and Shorel were both right — added it. Thanks -- the lexicographers like https://www.sketchengine.eu

Fair point — added a mention of collocations

I'm the author. Updated the article based on this thread — thanks to everyone who pushed back. Changes: reframed "boiling water" as deliberately on-the-fence rather than asserting it's a word; added a note that the obscure end of the slider is noise; acknowledged collocations as the established term; added a German/Dutch/Norwegian section on how other languages handle the space problem; softened "wasn't possible before LLMs" to "wasn't practical"; and threaded the concept of "loaded" throughout as the key distinction. Many of the specific examples came directly from commenters here — credited below.

The very bottom of the slider is there to illustrate where LLM artifacts and Wiktionary noise live — it's not presented as legitimate vocabulary. The slider lets you see the full quality gradient, including where it breaks down.

Fair point. I just rewrote the intro w/ the naming-function argument first.

  Location: Takoma Park, MD, USA
  Remote: Yes
  Willing to relocate: No
  Technologies: Python, Swift, JavaScript, semantic systems, LLM orchestration
  Portfolio: https://www.michaeldouma.com
  Email: m5@inotherwords-studio.com

I make sophisticated systems feel intuitive.

Track record: time.gov (1999, still running), SpicyNodes (40M users), ProstateCalculator (1.5M patients navigating treatment decisions). Last decade running a word games studio built on a semantic knowledge graph—120M+ connections across 1.7M words, NSF funded ($295K SBIR).

Game design taught me what most product teams skip: progressive disclosure, intrinsic motivation, making the 100th use as rewarding as the first. I apply that to software that isn't games. Judge for IGF, MagFest, CODiE, GEE, Serious Games.

Looking for: Product/design leadership at Series A–C companies (10–50 people). Developer tools, learning platforms, knowledge products, brain health, creative AI. Teams where I can lead and ship, not just roadmap. Products that need to feel less like enterprise Java circa 2003.

  Latest: https://www.inotherwords.app
  LinkedIn: https://www.linkedin.com/in/michaeldouma/

This is a follow-on piece to “The Small World of English” (at inotherwords.app)

There are nearly half a million compound phrases that aren’t in any dictionary—simply because they contain spaces. “Boiling water.” “Saturday night.” “Help me.” I got interested in this because I make word games. I wanted to understand when to include words with spaces—and the legacy effects of traditional dictionaries on our sense of “what is a word.”