00 · edelbyte.io, first response
What would an AI agent learn from edelbyte.io?
> agent.fetch("https://edelbyte.io")
← 200 OK · 4,536 bytes · 7 words
Edelbyte: The data platform to build on
That’s all of it. The rest of your homepage arrives with JavaScript, and most AI crawlers don’t run JavaScript.
- In the HTML · agents can read it
- After JavaScript · agents see nothing
Same page. Two readers.
✓ MD5 identicaldc18626a80d4bfe9eceb967b148879f8GPTBot receives these exact 4,536 bytes, as do the other three.
Hover to move the lens. Keyboard: Tab to it, arrows move, Enter switches reader.
Why now
- NOV 11
You came out of stealth with $70M. The coverage is out there.
- NOW
Readers ask ChatGPT, Claude and Perplexity “What is Edelbyte?” and “Edelbyte vs Snowflake?”
- NOV 30
re:Invent. Thousands more people will ask the same questions.
Those answers are written from whatever text the crawlers could read. Right now that’s other people’s articles, not your page.
Four questions. Two answers each.
Your homepage animates a query becoming an answer. Here are four an agent would ask, answered from the HTML it receives and from the page people see.
What is Edelbyte?
0 matchesfor “any data challenge” in Edelbyte: The data platform to build on
Edelbyte is the unified data platform for any data challenge, from search and analytics to AI applications…
Can it run on-premises?
0 matchesfor “on-premises” in Edelbyte: The data platform to build on
Run in any cloud, self-hosted, or on-premises.
Does it replace separate databases?
0 matchesfor “specialized databases” in Edelbyte: The data platform to build on
Search, analytical queries, vector, graph, geospatial and more, all in one platform. No need to stitch together specialized databases.
How fast is it at scale?
0 matchesfor “petabytes” in Edelbyte: The data platform to build on
Milliseconds on petabytes.
answered from HTML: 0 / 4·answered for people: 4 / 4
Every URL returns the homepage.
/
<title>Edelbyte: The data platform to build on
canonical → /200 · 4,536 B · md5 dc18626a…
/about
<title>Edelbyte: The data platform to build on
canonical → /200 · 4,536 B · md5 dc18626a…
/pricing
<title>Edelbyte: The data platform to build on
canonical → /200 · 4,536 B · md5 dc18626a…
/contact
<title>Edelbyte: The data platform to build on
canonical → /200 · 4,536 B · md5 dc18626a…
/this-page-does-not-exist
<title>Edelbyte: The data platform to build on
canonical → /200 · 4,536 B · md5 dc18626a…
To a crawler these are one page. There are no real 404s, and /about declares itself canonical to /. That one matters for Google too.
UA="Mozilla/5.0 (compatible; GPTBot/1.2; +https://openai.com/gptbot)" curl -sL -A "$UA" https://edelbyte.io -o home.html -w "%{http_code} · %{size_download} B\n" 200 · 4536 B md5 -q home.html # md5sum on Linux dc18626a80d4bfe9eceb967b148879f8 # same for ClaudeBot, PerplexityBot, Chrome grep -ao "<title>[^<]*</title>" home.html <title>Edelbyte: The data platform to build on</title> sed -E 's#<script[^>]*>[^<]*</script>##g; s#<[^>]+># #g' home.html | wc -w 7 grep -c "Milliseconds on petabytes" home.html 0
Run it yourself. Ten seconds.
Already right
- meta description
- Organization JSON-LD
- /llms.txt
- /sitemap.xml
The groundwork is there. The pages it points to just need to return their own content.
What to change, and how to check it.
Prerender the marketing routes
TanStack Start prerender or SSR (check against your version). The body copy ships in the HTML.
curl -sA GPTBot https://edelbyte.io | grep -c "Milliseconds on petabytes" → 1
A head of its own for every route
Each route gets its own title, description and canonical. Unknown paths return 404.
curl -o /dev/null -w "%{http_code}" https://edelbyte.io/nope → 404
Extend the structured data
Add SoftwareApplication and Person schema next to the existing Organization.
Rich Results Test → passes
Make llms.txt point at readable pages
Every page listed in llms.txt returns readable text, optionally with .md twins.
curl -sA GPTBot https://edelbyte.io/about | sed -E 's/<[^>]+>/ /g' | wc -w → > 200
Pages built to be quoted
Edelbyte vs a stitched stack, Architecture, Use cases.
AI answers cite edelbyte.io
Guard it in CI
A check that fails the build if the homepage HTML loses its headline.
grep -q "What would you build" dist/index.html || exit 1
None of this changes the design you shipped. It changes what the first response contains.
Data where it lives.
An architecture section built from real HTML text, so a crawler reads the same diagram a person does. Hover or focus a workload to trace its path.
Your data
- Apache Iceberg
- Parquet
- S3
- Kafka
- Delta Lake
- PostgreSQL
- MySQL
Edelbyte
Tools & agents
- Claude
- Codex
- GitHub
- Slack
- Grafana
- dbt
- Looker
- Tableau
- Power BI
- Jupyter
“Find every incident ticket that mentions checkout latency this week.”
“Revenue by region for the last 90 days, refreshed every minute.”
“Which past support answers are closest to this customer’s question?”
“Which services depend on the payments database, two hops out?”
“Why did p99 latency spike at 14:02, and what changed?”
Same query, after the fix.
- Today · measured
- 4,536 B · 7 words
- Target · not measured
- ~40 KB · 600+ words
> agent.fetch("https://edelbyte.io")
← 200 OK · ~40 KB · 600+ words
Edelbyte is the unified data platform for any data challenge, from search and analytics to AI applications…
Shivraj Nag · Mad Method · madmethod.io
We build product sites for technical companies and get them read by the engines people now ask.