I’ve long been impressed with the performance of the search in Metaphor Systems, from taking long-form queries and autoprompting, to allowing a range of filtering options and finding similar pages.1 I was recently trying the similar search again, reflecting on Metaphor’s “Job Search with Metaphor” showcase, written by Sarah Chieng. In that post she explains how to write a Metaphor query, then describe the similar search function:
But, there’s actually an even better ✨magical✨ way to take advantage of Metaphor. I can just paste a job posting I like and get similar ones.
Rather than paste a job posting, I pasted in my hire-me page (which was indexed by Metaphor), excluded my own website and filtered to just the past month:
I was floored, the results were very relevant, including pages and posts I had already seen and liked. I decided to experiment with the Metaphor API by writing up a simple script I could use to periodically run a few of these find_similar searches. This filters to exclude the domain of the searched-URL (using exclude_source_domain
) and filters to only pages published within the last 30 days (start_published_date
).2
metaphor-python
I have several pages now set to be monitored and logged:
The results include positions, people, and pages I’ve already known about and now ones I look forward to looking into. Relevance isn’t perfect, but the experience is rather like serendipity. (Note that relevance is constrained by the date filter. Experiment with different settings for your own searches and search purposes.)
Here is a display of my most recent search:
No Title
URL: https://twitter.com/profjamesevans/status/1715411595554402345
Director of Product, Firefox Search and AI
Assistant Professor - Information Science
URL: https://jobs.colorado.edu/jobs/JobDetail/Assistant-Professor-Information-Science/52624
No Title
URL: https://twitter.com/ej_mcdonnell/status/1716897717551804521
Request for proposals: studying and forecasting the real-world impacts of systems built from LLMs | Open Philanthropy
No Title
No Title
URL: https://twitter.com/PrivacyLawyerD/status/1724244681028940137
Data Scientist II
Cornell University, Information Science
We’re Hiring: Spring 2024 Google U.S. Public Policy Fellowship Program
URL: https://techfreedom.org/were-hiring-spring-2024-google-u-s-public-policy-fellowship-program/
Digital Theory Lab
URL: https://digitalhumanities.nyu.edu/projects/digital-theory-lab/
Lindsay Brainard
Home
Doc Searls Weblog | Old blog, new place
UX Roundup: OpenAI & GitHub AI Announcements | Jakob’s Razor for UX Research | Midjourney Beta | AI Censorship | Anthropomorphizing AI
URL: https://www.linkedin.com/in/jakobnielsenphd/?original_referer=
Megan Marz
half thoughts
L.M. Sacasas
E. J. Green - Research
Wrench in the Gears
Andrew Reddie
E. J. Green
Mr. Harris goes to Brussels (Adventures in AI Policy Part II)
URL: https://www.linkedin.com/in/davidevanharris/?original_referer=
Megan G. Massa, PhD
UX Roundup: OpenAI & GitHub AI Announcements | Jakob’s Razor for UX Research | Midjourney Beta | AI Censorship | Anthropomorphizing AI
URL: https://www.linkedin.com/in/jakobnielsenphd/?original_referer=
Megan Marz
Lindsay Brainard
far.in.net
URL: https://far.in.net
Zhengyi Zhou
Joshua Siegel | College of Engineering
Observers Observed: The Ethnographer in Silicon Valley
URL: https://crookedtimber.org/2023/11/09/observers-observed-the-ethnographer-in-silicon-valley/
On building a home on the web • Daniël van der Winden
Examining bias perpetuation in academic search engines: an algorithm audit of Google and Semantic Scholar
The 4+1 Model of Data Science
Data Science: The Hard Parts
Contextualizing the Limits of Model & Evaluation Dataset Curation on Semantic Similarity Classification Tasks
Navigating Complex Search Tasks with AI Copilots
Data Enclaves
URL: https://link.springer.com/book/10.1007/978-3-031-46402-7
Human participants in AI research: Ethics and transparency in practice
Education for a Future in Crisis: Developing a Humanities-Informed STEM Curriculum
Home
Mr. Harris goes to Brussels (Adventures in AI Policy Part II)
URL: https://www.linkedin.com/in/davidevanharris/?original_referer=
Google Scholar Henvisninger
Google Scholar Citations
Google Scholar Citations
E. J. Green - Research
Google Scholar Citations
Zhengyi Zhou
Karthik Srinivasan
maya weeks
I was compelled to finally try out the Metaphor API a bit more as I explored the GPTs that Jeffrey Wang prepared:
If you’re building a dev-facing product, I highly recommend making GPTs that answer questions about it! Some benefits: - Can explain product nuances - Can write code - Familiar form factor
Below are some GPTs I built for Metaphor Search API. Lemme know what you think!
Google has had “Pages similar to:” on its search results page. (Though I don’t think it is currently provided?) See this post: “Discovering pages “similar to” ones that you like" (Apr 27, 2010; by Doantam Phan)
↩︎We’ve offered a “Similar” feature on results for a while now as a way to discover new, useful sites, but it hasn’t been too visible. Since we’ve been continuously improving this feature and we think it’s really useful, we’re now going to start showing these alternative sites more prominently. Starting this week, for queries where similar sites are likely to be helpful, we’ll display a list of “Pages similar” at the bottom of the results page.
The date filter likely uses the Last-Modified HTTP header (which is editable by the publisher) so it does not indicate when a page was created, but hopefully does indicate when substantive content changes were made.↩︎