Runs September 1, 2026 8:00 AM EST to September 3, 2026 8:00 AM EST
(12:00 UTC to 12:00 UTC), linking to the launch blog post.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This event covers machine learning, computational biology, and lab automation. It aims to inform researchers and engineers about current capabilities and challenges in drug discovery.
* Apply the quote component to OpenTable, &AI and Dust
Seven quotes, all verbatim. Only quotes that say something about Qdrant
were lifted; setup and problem-statement quotes stay in the prose they
were written for.
OpenTable has three different speakers, so each card names its own.
Two &AI quotes were split across "said Turner" in the original. Each card
carries only the contiguous half about Qdrant, and the setup fragment
stays in the prose - joining the halves would have meant editing his
words, even though every word was his.
Dust carried the same quote twice: once inside the alt text of
Dust-Quote.jpg and again as prose below it. The image is gone and the
words are a card, so the quote is now text rather than pixels and is no
longer duplicated.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Add a quote-finding script, and promote two lead quotes
Earlier passes over these files missed quotes because the search required
an attribution verb near the quotation. Anything attributed on its own
line with a dash, standing alone as an italic pull quote, sitting in a
blockquote, or buried in an image's alt text was never surfaced.
automation/find-customer-quotes.py reports all of those forms, and
deliberately over-captures: it flags any quoted run of 25 characters or
more outside front matter, code fences and existing shortcodes. A false
positive costs a moment's reading; a missed quote ships. Across the 52
case studies it finds 201 quotes still in prose.
Also promotes the strongest quote in two files to a lead card, taking the
logo with it so it stays on the first card on the page:
- opentable: "Creating a Qdrant Cloud cluster was one of the easiest
parts of the project. It just worked."
- and-ai: "With Qdrant, we scaled to a billion vectors and still respond
in sub-second latency."
Dust is left alone. Its one card follows a sentence written to introduce
it, so moving it would orphan that sentence.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Move the OpenTable lead quote to one that carries its own context
"Creating a Qdrant Cloud cluster was one of the easiest parts of the
project" was sitting above every line of prose in the article. It only
lands once you know there was a project and that it had hard parts, so as
an opener it read as a punchline with no setup. It goes back after "Third,
Qdrant Cloud provided a deployment path that was simpler than
self-hosting", which is the context it was written against.
The article already had a Lombardo quote near the top explaining why
OpenTable built Concierge at all. That one needs nothing before it, and it
already sat after the opening paragraph, so it becomes the lead card
without being moved anywhere.
A lead quote has to make sense to someone who has read only the title, and
it belongs after the opening paragraph rather than above it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Show the customer logo on every quote card
The logo used to sit only on the first card in each article, so later
quotes carried no visual marker of who was speaking beyond the name and
role. Adds it to the remaining 18 cards: 10 in Bayer, 3 each in OpenTable
and Sprinklr, 2 in &AI.
Note this hides the company name on those cards, since the component drops
it when a logo is present. The name still appears on narrow screens, where
the logo is hidden.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Convert the quotes still left in prose in &AI and Dust
Working through what find-customer-quotes.py reported rather than stopping
at the first few. &AI goes from 3 cards to 8, Dust from 1 to 4.
Where the original splits a quotation across "said Turner", the card takes
only the contiguous half that says something about Qdrant and the setup
half stays in the prose, so nothing is joined and no words change.
What is left in those two files is deliberate: the setup halves, and
Polu's remarks about LLMs and fine-tuning in general, which are context
rather than anything about Qdrant.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Drop the quote-finding script from the branch
Keeping it out of the repo. It was only ever scaffolding for this
conversion work, not something the site needs to build.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Spend the avatar and logo only on the quotes worth marking out
Putting them on every card spent them on none - the face and the logo are
what make a card read as a headline, so they only work if most cards go
without.
Bayer goes back to how it was on master: two marked out of twelve.
Each of the others keeps them on the one or two quotes making the
strongest claim:
- opentable: the lead, and "Creating a Qdrant Cloud cluster was one of the
easiest parts of the project. It just worked."
- and-ai: the lead, and "Qdrant was the only one that handled that load
out of the box."
- dust: "we were able to reduce our cost by 2x", the one with a number.
- sprinklr: "Retrieval is the foundation of all our AI tasks."
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Dust and Sprinklr fixes from review
Two quotes go back into the prose they came from. "The early setup worked
out of the box nicely" sits inside a numbered list item, which is the
in-bullet case that should have been left alone. "Qdrant is a very fast
and high quality retrieval system" does not say much the surrounding
benchmark numbers have not already said. Both restored exactly.
Deletes Dust-Quote.jpg, now that nothing references it, and bolds "reduce
our cost by 2x" in the quote that reports it.
Embeds the Vector Space Talk with Stanislas Polu after the paragraph
linking to it.
Adds Raghav Sonavane's photo, supplied by the team, cropped to head and
shoulders like the other avatars. It goes on the one Sprinklr quote that
carries the logo.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Rejoin the &AI quotes that were split across an attribution tag
Each of these is one statement in the original, interrupted only by "said
Turner". I had put the second half in the card and left the first stranded
in the prose, so the card carried half a thought and the setup sat above it
going nowhere.
The tag goes, the comma that introduced it becomes a period, and the two
halves join. No words change - checked that both halves still appear
verbatim in the original either side of the join.
Affects five quotes: retrieval over generation, the scale benchmark, the
opinionated docs, multi-tenancy, and the billion-dollar case.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Clarify tiered 1-shard limitation and update code snippets to show best practice
* Split write/read multitenancy section
* Use consistent example throughout the page
* Hide client initialization from all code snippets
kanungle