Founder @And/or Labs; building intelligence products for media, adtech, and emerging tech.
Calgary, Alberta
Joined February 2009
- Tweets151K
- Following1.6K
- Followers3.9K
- Likes8.2K
On a recent demo call, a potential buyer was trying to grill me about how I’m sourcing such an extensive collection of data about media/publishing entities.
Bro, if I told you that, you’re just gonna use that as evidence that “anyone can build this”. Go do your own research.
Of course, while the WebKit team kneecaps adtech, Apple has been building personal graphs on all their users since 2022.
Apparently all the brand advertising about them being the vanguards of privacy just meant: “We won’t let anyone else do it.”
Building software for internal ops is a distraction from the main product you’re building as well as a massive time sink—terrible idea all things considered. This is why I find the “I built my own CRM” crowd to be a bunch of unserious jokers.
One it the consequences of the new prompt-to-build-anything world is the fact that you can technically get things done without the educational benefits of doing anything even lightly graze you.
This here is a good example. The majority of people who don’t know what implicit state machines are will make the fix and move on to the next thing. Or at least that’s what I think.
Justin, Drake, and The Weeknd are gonna have a fucking banner year.
I’m so tired of staring at terminals all day. The last felt this type of tiredness was troubleshooting BSOD back when MSFT couldn’t ship a stable OS if their life depended on it.
Vishveshwar Jatain retweeted
To all the AEO/GEO tracking software that raised hundreds of millions:
Free tracking inbound from @OpenAI
Vishveshwar Jatain retweeted
made 40 GTM brains with Pancake
copy a brain and give that context to your agent
gtmbrain.si
our research on how context affects GTM recommendations below
Sucks for all the agent meta harnesses, but I actually enjoy working out of 99 parallel terminal tabs.
How easily can AI agents find and use your product?
These benchmarks help builders find the right infrastructure blocks faster and help product companies improve their adoption by agents.
I used @neondatabase for my product and not surprised to see it rank high, as one of my key product selection criteria was hands-off MCP connection. I’ve probably only logged in their UI twice.
Definitely recommend following @tryscope_app for more.
🤝 Paid partnership
How much revenue are database companies losing because agents aren’t able to sign up autonomously?
Today, we’re releasing @tryscope_app's first database benchmark.
We tested 18 hosted databases to see which are ready to win that market. We measured which databases agents could use, analyzed where they failed, and identified best practices to fix the blockers.
Across 1,035 sessions, only 16.6% reached a first working result and failed sessions all came from products not being agent friendly.
We tested 20 model-and-agent-tool combinations across harnesses including Claude Code, Codex and Cursor.
The task: start with no account, set up the named database, write a record, and hand over the credentials. An account or API key alone didn’t count.
What we found:
→ @prisma led with an 88.3% success score, followed by @upstash at 63.3%, @neondatabase and @niledatabase at 43.3%, and @ClickHouseDB at 38.3%.
→ 7 of 18 products had zero successful sessions across every agent tested.
→ CAPTCHAs stopped 373 sessions. Bot blocks stopped another 105. Other barriers included rejected agent email domains and OAuth-only flows.
→ 50 of the 172 successful sessions used a working database without creating an account. Letting agents try the product before signing up made a difference.
An agent can arrive knowing exactly which product it wants to use and still never become a user. For developer tools, the path from “use this database” to a working result is part of your conversion funnel.
See the full leaderboard, methodology and session transcripts: tryscope.com/leaderboard/dat…
Special thank you to our partners @AgentMail and @Agentphone for giving us the right tools to run this experiment.
Told my 2yo that I’ll get her a circuit board to play with and now every time she sees me, she points a finger up and says, “Where’s my circuit board?”
“Can’t I just ask Claude the same question about a publisher?”
Of course. MediaContext is merely a persistently queryable, domain-specific relational information substrate with heterogeneous multi-source ingestion and bespoke classification orchestration.
Claude is an AI model. Practically twins, once you abstract away the architecture, function, and purpose. We ingest, normalize, classify, and establish relational and cosine similarity between records.
Could Claude access those records?
Absolutely. Just provision the ingestion pipelines, implement the persistence layer, configure the classifiers, expose the query interface, and integrate the model.
You’ve now recreated the app. A tremendous victory for not needing the app.
“But surely it’s all just AI?”
Yes. And a warehouse is just a forklift, if you bring your own building, inventory, shelving, and increasingly exhausted understanding of what a forklift is.
We’re not replacing generative intelligence.
We’re operationalizing the upstream contextual prerequisites for downstream answer relevance.
We’re putting information in a database.
But vertically.
