<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: RobinL</title><link>https://news.ycombinator.com/user?id=RobinL</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Thu, 19 Nov 2026 15:08:50 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=RobinL" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by RobinL in "GPT‑6.2 Astra and Sol"]]></title><description><![CDATA[
<p>Ethan Mollick too:
<a href="https://bsky.app/profile/emollick.bsky.social/post/3n2joens5t22x" rel="nofollow">https://bsky.app/profile/emollick.bsky.social/post/3n2joens5...</a></p>
]]></description><pubDate>Tue, 17 Nov 2026 18:53:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=50536351</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50536351</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50536351</guid></item><item><title><![CDATA[New comment by RobinL in "Falsehoods programmers believe about addresses (2013)"]]></title><description><![CDATA[
<p>What a classic!  I work on a free address matching solution and this article was formative in deciding on the approach.<p>In particular, it's common for address matching solutions to begin by parsing the address into it's constituent parts (number, street name etc.).  But articles like this suggested to me this is misguided: it works fine for easy to match addresses, but the addresses which are hard to match are hard to parse, and hence the parsing problem and matching problem reduce to the same problem.<p>Instead, many of the features that are useful for matching addresses are more related to looking at the address in the context of the other addresses around it and in the corpus, rather than looking just at the address itself.  I've written a bit about this here: <a href="https://www.robinlinacre.com/address_matching/" rel="nofollow">https://www.robinlinacre.com/address_matching/</a></p>
]]></description><pubDate>Tue, 10 Nov 2026 12:01:37 +0000</pubDate><link>https://news.ycombinator.com/item?id=50439100</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50439100</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50439100</guid></item><item><title><![CDATA[New comment by RobinL in "DuckDB 2.0.0 LTS “Cinnamon Teal”"]]></title><description><![CDATA[
<p>Will try this thanks!</p>
]]></description><pubDate>Thu, 22 Oct 2026 06:36:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=50191967</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50191967</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50191967</guid></item><item><title><![CDATA[New comment by RobinL in "DuckDB 2.0.0 LTS “Cinnamon Teal”"]]></title><description><![CDATA[
<p>Yep that makes sense.<p>One thing we've seen with DuckDB, which they also mention in their docs, is that it's possible to 'over parallelize'.  Beyond some number of cores, the overall workflow measured in wall time actually slows down when you add more CPUs, because of the overhead of parallelizing, despite them running at 100%.<p>We observe this even if we pull the data from S3 onto the local SSD before processing,</p>
]]></description><pubDate>Wed, 21 Oct 2026 17:29:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=50184587</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50184587</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50184587</guid></item><item><title><![CDATA[New comment by RobinL in "DuckDB 2.0.0 LTS “Cinnamon Teal”"]]></title><description><![CDATA[
<p>Appreciate the kind words.<p>There are a variety of people still use Spark, but this is usually due to constraints on their platform/infrastructure, not because it performs better.  For instance people on Databricks can sometimes find it hard to use DuckDB on large machines, or they're just more familiar with Spark.<p>For a new user who has flexibility over where to run their workloads, we'd always recommend DuckDB.  This is particularly the case now that:
- Splink 5 (released in September) allows you to run your workload in n chunks, meaning you can distribute it across multiple machines even on DuckDB
- DuckDB 2.0 parallelises better on large machines, so throwing a single massive machine at problems is now more viable<p>In general, after using Spark for a long time, I think for most workloads, it's just not really needed any more.  Too complex, too difficult to tune.  Perhaps if you have genuinely massive data.  I've written a bit more about why I advise against Spark in most instances here: <a href="https://www.robinlinacre.com/recommend_duckdb/" rel="nofollow">https://www.robinlinacre.com/recommend_duckdb/</a></p>
]]></description><pubDate>Wed, 21 Oct 2026 15:35:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=50182776</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50182776</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50182776</guid></item><item><title><![CDATA[New comment by RobinL in "DuckDB 2.0.0 LTS “Cinnamon Teal”"]]></title><description><![CDATA[
<p>We've found DuckDB 2.0 to be way faster on high core count machines on EC2.  Not just reading data from S3, but the actual computations too. Much more likely to see 100% utilisation up to about 128vCPUs or 192vCPUs, and that CPU usage to translate into workloads completing faster.<p>Still don't usually see a speedup  past about 192vCPU, but prior to 2.0, speed would top out at more like 64vCPU, so this is a big win.<p>That has enabled us to scale Splink (FOSS record linakge software) to process 1bn records with 10bn record comparisons in 10 mins, which as far as I'm aware makes it the fastest linkage engine of this kind. Most of this performance comes from simply delegating the big calculations to the DuckDB engine!  <a href="https://moj-analytical-services.github.io/splink/topic_guides/performance/duckdb_performance_reprex.html" rel="nofollow">https://moj-analytical-services.github.io/splink/topic_guide...</a><p>Thanks to the team for all their work!</p>
]]></description><pubDate>Wed, 21 Oct 2026 14:27:51 +0000</pubDate><link>https://news.ycombinator.com/item?id=50181617</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50181617</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50181617</guid></item><item><title><![CDATA[New comment by RobinL in "The United Kingdom generates less electricity per person than in the 1970s"]]></title><description><![CDATA[
<p>The spot price is set according to the most expensive source but that is not the same thing as the price the consumer pays.<p>So for example, for wind power the consumer effectively pays the CfD strike price, largely irrespective of the spot price.<p>Fwiw, I do agree we have a problem with power being too expensive, but it's more complex than just the consumer pays the most expensive source.</p>
]]></description><pubDate>Tue, 13 Oct 2026 21:31:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=50079105</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=50079105</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50079105</guid></item><item><title><![CDATA[Splink 5.0.0 Released (probabilistic record linkage at scale)]]></title><description><![CDATA[
<p>Article URL: <a href="https://moj-analytical-services.github.io/splink/blog/2026/09/28/splink-500-released.html">https://moj-analytical-services.github.io/splink/blog/2026/09/28/splink-500-released.html</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49891072">https://news.ycombinator.com/item?id=49891072</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 29 Sep 2026 10:53:18 +0000</pubDate><link>https://moj-analytical-services.github.io/splink/blog/2026/09/28/splink-500-released.html</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49891072</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49891072</guid></item><item><title><![CDATA[Why parquet files are my preferred API for bulk open data (2023)]]></title><description><![CDATA[
<p>Article URL: <a href="https://www.robinlinacre.com/parquet_api/">https://www.robinlinacre.com/parquet_api/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49863888">https://news.ycombinator.com/item?id=49863888</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sun, 27 Sep 2026 06:23:40 +0000</pubDate><link>https://www.robinlinacre.com/parquet_api/</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49863888</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49863888</guid></item><item><title><![CDATA[New comment by RobinL in "Introducing System One Models and Jev"]]></title><description><![CDATA[
<p>I agree, and I think LLMs can potentially do a better job than more conventional methods, so long as they're provided with enough content.<p>Another aspect of context that could be relevant to your work may be how people are nested within households.  In some record linkage scenarios you can run a linkage on household membership and use this to make person linkages more precise.  I imagine an LLM agent could also use this information effectively if it were explicitly provided, but sometimes they miss logical leaps like this</p>
]]></description><pubDate>Wed, 16 Sep 2026 11:17:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=49724902</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49724902</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49724902</guid></item><item><title><![CDATA[New comment by RobinL in "Introducing System One Models and Jev"]]></title><description><![CDATA[
<p>There's also two other important limitations to using an LLM and just providing it with pairs of records.<p>It does not know enough about the records in the context of the overall dataset:<p>- what is the data quality and to what extent do we expect a errors in some fields
- how unusual are certain values such as names in the context of the dataset as a whole, e.g. some names would be very common in some countries but rare in others.<p>I've written in more detail about this here: <a href="https://www.robinlinacre.com/fellegi_sunter_accuracy/" rel="nofollow">https://www.robinlinacre.com/fellegi_sunter_accuracy/</a></p>
]]></description><pubDate>Wed, 16 Sep 2026 10:43:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=49724538</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49724538</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49724538</guid></item><item><title><![CDATA[New comment by RobinL in "Show HN: An e-ink frame that hears birds and draws them as 1800s illustrations"]]></title><description><![CDATA[
<p>> Half the point of this project is showing off some amazing public-domain natural-history illustrations. Over 800 cut-outs covering more than 400 species, every one taken from a real plate and hand-curated for this project (no art is AI-generated, though some has been retouched with AI).</p>
]]></description><pubDate>Tue, 15 Sep 2026 14:16:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=49712883</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49712883</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49712883</guid></item><item><title><![CDATA[New comment by RobinL in "Ask HN: What are you working on? (September 2026)"]]></title><description><![CDATA[
<p>Really like this, well done!</p>
]]></description><pubDate>Mon, 14 Sep 2026 06:07:31 +0000</pubDate><link>https://news.ycombinator.com/item?id=49692602</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49692602</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49692602</guid></item><item><title><![CDATA[New comment by RobinL in "Ask HN: What are you working on? (September 2026)"]]></title><description><![CDATA[
<p>I've been experimenting for a year or two making games that make mental arithmetic practice fun for kids.  The latest is
<a href="https://rupertlinacre.com/keep_it_going/" rel="nofollow">https://rupertlinacre.com/keep_it_going/</a> an infinite rollercoaster, which I think is one of the more successful attempts.  There are various others on the homepage. My son also really likes arithmetic annihilation and maths vs monsters.</p>
]]></description><pubDate>Mon, 14 Sep 2026 05:52:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=49692497</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49692497</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49692497</guid></item><item><title><![CDATA[A New Observable]]></title><description><![CDATA[
<p>Article URL: <a href="https://new.observablehq.com/@observablehq/a-new-observable">https://new.observablehq.com/@observablehq/a-new-observable</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49629276">https://news.ycombinator.com/item?id=49629276</a></p>
<p>Points: 4</p>
<p># Comments: 0</p>
]]></description><pubDate>Wed, 09 Sep 2026 16:37:23 +0000</pubDate><link>https://new.observablehq.com/@observablehq/a-new-observable</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49629276</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49629276</guid></item><item><title><![CDATA[New comment by RobinL in "ChatGPT Images 2.5"]]></title><description><![CDATA[
<p>I wish I could do it better. There seems to be a lot of variability.  This episode of radiolab on the topic is fascinating: <a href="https://radiolab.org/podcast/aphantasia/transcript?utm_source=chatgpt.com" rel="nofollow">https://radiolab.org/podcast/aphantasia/transcript?utm_sourc...</a></p>
]]></description><pubDate>Tue, 08 Sep 2026 19:29:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=49615618</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49615618</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49615618</guid></item><item><title><![CDATA[New comment by RobinL in "ChatGPT Images 2.5"]]></title><description><![CDATA[
<p>Interesting. I'm surprised they haven't prioritised/deliberately trained for this more because of how useful it is to ask codex to generate some images/assets/sprites when building sill games.  It would dramatically enhance how polished it's games were.<p>That said for single images the old model was already okayish for prototypes</p>
]]></description><pubDate>Tue, 08 Sep 2026 19:14:10 +0000</pubDate><link>https://news.ycombinator.com/item?id=49615332</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49615332</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49615332</guid></item><item><title><![CDATA[Data Science in the Age of AI]]></title><description><![CDATA[
<p>Article URL: <a href="https://www.robinlinacre.com/data_science_age_ai/">https://www.robinlinacre.com/data_science_age_ai/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49519172">https://news.ycombinator.com/item?id=49519172</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 01 Sep 2026 07:41:57 +0000</pubDate><link>https://www.robinlinacre.com/data_science_age_ai/</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49519172</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49519172</guid></item><item><title><![CDATA[New comment by RobinL in "Show HN: Buslens – where can I get to by bus? (UK)"]]></title><description><![CDATA[
<p>Thanks - author here - I think this is due to the unexpected traffic spike from Hacker News!  Will look into</p>
]]></description><pubDate>Wed, 26 Aug 2026 14:56:33 +0000</pubDate><link>https://news.ycombinator.com/item?id=49450301</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49450301</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49450301</guid></item><item><title><![CDATA[New comment by RobinL in "Show HN: Buslens – where can I get to by bus? (UK)"]]></title><description><![CDATA[
<p>Thank you!  My son's favourite at the moment is arithmetic annihilation in two player mode:
<a href="https://rupertlinacre.com/arithmetic_annihilation/" rel="nofollow">https://rupertlinacre.com/arithmetic_annihilation/</a></p>
]]></description><pubDate>Wed, 26 Aug 2026 12:33:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=49447889</link><dc:creator>RobinL</dc:creator><comments>https://news.ycombinator.com/item?id=49447889</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49447889</guid></item></channel></rss>