Top all-time · #2288
Doug Turnbull
Personal blog · softwaredoug.com · on HN as softwaredougaidataprogramming
Doug Turnbull's personal blog at softwaredoug.com covers ai, data and programming. Since 2019, 35 different Hacker News readers have linked it 229 times, most often “What AI Engineers Should Know about Search”. It is still writing: the last post was on 2026-10-26, and you can follow it by feed.
Most-linked posts?
- What AI Engineers Should Know about Search4 people · 4 mentions · first linked 2024-06-25
- Coping with dumb LLMs using classic ML3 people · 4 mentions · first linked 2025-01-22
- Turning my laptop into a Search Relevance Judge with local LLMs3 people · 3 mentions · first linked 2025-01-13
- Staff engineers exist in a system of patronage3 people · 3 mentions · first linked 2024-09-11
- Elasticsearch Hybrid Search in Practice3 people · 3 mentions · first linked 2025-02-10
- Don't classify, hallucinate!3 people · 3 mentions · first linked 2026-08-10
- Dont have F-You money? Build an F-You Network2 people · 3 mentions · first linked 2024-05-08
- OpenAI lost the plot on boring LLM use cases (2025)2 people · 3 mentions · first linked 2025-08-21
- RRF Is Not Enough2 people · 3 mentions · first linked 2024-11-04
- LSH in NumPy C for some fun but no profit2 people · 3 mentions · first linked 2023-08-23
What Hacker News says?
https://softwaredoug.com/blog/2026/08/10/hypothetical-classi...
Comment linking /blog/2026/08/10/hypothetical-classifications
on “Introducing System One Models and Jev”Submitted Idiot Proof Git as a story · 387 points
HN discussionYeah that's precisely correct. For e.g. classification tasks, even in 2026 people are doing things like hallucinating deliberately, and then matching the hallucinations to embeddings - https://softwaredoug.com/blog/2026/08/10/hypothetical-classi ... With TypeSafe it just picks the class (actually probabilities across classes), reliably every single time.
Comment linking /blog/2026/08/10/hypothetical-classi
on “Introducing System One Models and Jev”Submitted Ruby vs. Python comes down to the for loop as a story · 373 points
HN discussionNothing, but imagine using LLMs for a classification task People out there are so resigned to the models being unreliable that they are really doing things like hallucinating deliberately, and then matching the hallucinations to embeddings - https://softwaredoug.com/blog/2026/08/10/hypothetical-classi... You could do that or you could just... use a model that will never produce unreliable outputs in the first place.
Comment linking /blog/2026/08/10/hypothetical-classifications
on “Introducing System One Models and Jev”Submitted Preferring throwaway code over design docs as a story · 327 points
HN discussionhttp://softwaredoug.com I write mostly about search, search relevance, etc
Comment linking /
on “Ask HN: Share your personal site”Submitted Are we at peak vector database? as a story · 235 points
HN discussionIt matches the training data. Whether the training data matches truth (and whether it's correctly understood - sarcasm included) is a completely separate thing. > The training does not perfectly align the model with truth, but 'orthogonal' Nitpicky, but the more dimensions you have, the easier it is for almost everything to be orthogonal. ( https://softwaredoug.com/blog/2022/12/26/surpries-at-hi-dime... ) That's why…
Comment linking /blog/2022/12/26/surpries-at-hi-dimensions-orthoginality
on “Differential Transformer”Submitted Coping with dumb LLMs using classic ML as a story · 184 points
HN discussionPretty cool! Recently I blogged how I make my twitter password really hard to recover through heavy, repeated hashing, which has helped mindless social media browsing https://softwaredoug.com/2020/04/05/kill-your-twitter-addict...
Comment linking /2020/04/05/kill-your-twitter-addiction.html
on “Show HN: I made a Chrome extension to stop mindless browsing”Submitted Ruby vs. Python comes down to the for loop (2021) as a story · 171 points
HN discussionI have vastly simplified my git workflow with some aliases that work in a variety of settings git synced #<- sync (rebase) current branch with upstream if defined, otherwise origin. Uses master or main, preferring main git pub #<- publish my changes to remote origin (force push with lease) git pr #<- open appropriate PR to github (no GH client needed, just opens the URL). PR goes to upstream if defined git hub #<- o…
Comment linking /blog/2022/11/09/idiot-proof-git-aliases.html
on “Git tips and tricks”Submitted Start with who, not why as a story · 160 points
HN discussionI agree with Simon’s article but I usually think about “research” to mean comparing different kinds of evidence (not just the search part). Like evidence for the effectiveness of Obamacare. Or how some legal case may play out in the courts. Or how much The Critic influenced The Family Guy. Or even what the best way to use X feature of Y library. I’ve found ChatGPT and other LLMS can struggle to evaluate evidence - t…
Comment linking /blog/2025/08/19/researching-with-agents
on “GPT-5 Thinking in ChatGPT (a.k.a. Research Goblin) is good at search”Submitted I made a search engine worse than Elasticsearch (2024) as a story · 141 points
HN discussionWhat software communities[1] are you a part of and who are you helping in those communities? Do you contribute there in the form of (code, documentation, blog posts, conference talks, helping juniors, etc)? At 40 I think you do well to have a decent network in a focused set of communities of people you've mentored, supported, and worked with. IE folks in a specific discipline, project, thing where you might expect t…
Comment linking /blog/2024/05/08/build-an-f-you-network
on “Ask HN: Where to Work After 40?”Submitted Companies need junior devs as a story · 130 points
HN discussionA warning that RRF is often not Enough, as it can just drag a good solution down towards the worse solution :) https://softwaredoug.com/blog/2024/11/03/rrf-is-not-enough
Comment linking /blog/2024/11/03/rrf-is-not-enough
on “Understanding the BM25 full text search algorithm”Submitted Bag of Decisions Reranker as a story · 75 points
HN discussionhttps://softwaredoug.com/blog/2022/12/26/surpries-at-hi-dime... it's both much more likely to be actually orthogonal and almost always very close to orthogonal.
Comment linking /blog/2022/12/26/surpries-at-hi-dimensions-orthoginality
on “Watch R1 "think" with animated chains of thought”Submitted Agents turn simple keyword search into compelling search experiences as a story · 67 points
HN discussionThe 2016 US election and me pouring endless energy into Twitter arguments with Trump supporters. Eventually too I read a lot of History about divisive times, and it showed me a lot of tragedy, much worse than what we're experiencing now. But also bounce back from that tragedy. You learn through life (and History) that Politics is relentlessly cyclical. Those that attempt control, even well intentioned, are ultimatel…
Comment linking /blog/2020/08/06/political-twitter-opposite-of-activism.html
on “Ask HN: What is it like to be old? What advice would you give to younger people?”Submitted Dont have F-You money? Build an F-You Network as a story · 58 points
HN discussionI wrote a blog about this actually - I also discovered it’s more valuable to enjoy who you’re working with, then what you’re working on https://softwaredoug.com/blog/2022/04/23/start-with-who.html
Comment linking /blog/2022/04/23/start-with-who.html
on “Ask HN: Thoughts on being “boring””Submitted An LLM Query Understanding Service as a story · 38 points
HN discussionI write for myself, and maybe a core group of friends, not random Web person. https://softwaredoug.com/blog/2024/05/16/just-code But this authenticity tends to make some new friends along the way.
Comment linking /blog/2024/05/16/just-code
on “Ask HN: Is maintaining a personal blog still worth it?”Submitted The tradeoff between human and AI context as a story · 34 points
HN discussionThis is a good opportunity to update my search reading list. Books * Lucene in Action - A bit out of date, but it has a lot of the core concepts of Lucene. 10 years ago, this was the go-to book even if you’re working with Solr or Elasticsearch, to understand the core data structures * Relevant Search (my book) - Introduction to optimizing the relevance of full text search engines. Getting a bit old now, but still re…
Comment linking /
on “Ask HN: I want to dive into how to make search engines”Submitted The other hard retrieval problems as a story · 26 points
HN discussion
