mritchie712

We started with a product like this at Definite (https://www.definite.app/), but it became clear there weren't enough people willing to spend real money on a product like it when Cursor / VS Code already have good coverage on data science.

he's also casually trashing their engineering team... "I built it in a day" is suggesting no one else could have.

I personally don't think there's anything wrong with doing it (i.e. shipping a feature quickly that a customer needs), but writing a blog post about it is questionable.

I'm with you. My favorite sci-fi is something like Project Hail Mary, which is radically different from fantasy.

DeepSeek released smallpond

0 - https://github.com/deepseek-ai/smallpond

1 - https://www.definite.app/blog/smallpond (overview for data engineers, practical application)

We started giving our (https://www.definite.app/) agent a sandbox (we use e2b.dev) and it's solved so many problems. It's created new problems, but net-net it's been a huge improvement.

Something like "where do we store temporary files the agent creates?" becomes obvious if you have a sandbox you can spin up and down in a couple seconds.

for many uses cases, you can cache a dataset client side that can be re-queried to avoid network calls

Yes, we run DuckDB + DuckLake in prod for https://www.definite.app/

Really excited about the future of DuckDB:

1. DuckLake is the best datalake spec and their team is improving on the extension rapidly.

2. With DuckDB WASM, you can make apps that would normally have 2 to 3 second latency for network calls work in < 200ms.

We use it as our built-in datalake at Definite and couldn't be happier with it.

0 - https://ducklake.select/

1 - https://www.definite.app/blog/ducklake

if you've ever worked with Excel + Python, I think this example will drive home the value a bit:

https://github.com/anthropics/skills/blob/main/document-skil...

There are many edge cases when writing / reading Excel files with Python and this nails many of them.

this is the best example I found

https://github.com/anthropics/skills/blob/main/document-skil...

I was dealing with 2 issues this morning getting Claude to produce a .xlsx that are covered in the doc above

It's the right move if they want to IPO.

The "modern data stack" market excluding of data warehouse / data lake is pretty small. Fivetran is biggest one and still under $500M in revenue, so they're acquiring other parts and starting to offer their own datalake (managed Iceberg).

Snowflake started offering fivetran-like connectors a couple years ago and I expect they'll double down there. Same with Databricks. Microsoft has Fabric now, but the reviews have been terrible (my experience included).

I think they'll each ultimately have a full data stack.

If you don't want to wait, we built a modern-data-stack-in-a-box at Definite (https://www.definite.app/).

they used GPT-4.1 nano, results would be quite different with sonnet or gpt5.

CC also has the same `/compact` command if you want to force it

do you have a video of someone using it?

Yeah, we're working on it. I'd prefer to by compute hours, but it's hard for people to know how many hours they'll need.

it'll be interesting to see what happens with n8n over the next few years.

On one hand, a ton of people are using it to build AI workflows.

On the other, it's never been easier to build a similar workflow in Python with Claude Code or other tools where you get infinite flexibility, version control, and more flexibility on hosting.

We're betting on the later at https://www.definite.app/. Our agent writes pure Python with pre-built integrations into stuff like Hubspot, Salesforce, Attio, Stripe, Postgres, etc.

huh? they can block the bots

how many GB does an app packaged with Qwen3 600m + Cactus take up?

e.g. if I built a basic LLM chat app with Qwen3 600m + Cactus, whats the total app size?

Have you used Claude Code? How does it compare?

For the comment above, the more relevant denominator is all humans vs. all developers. If you use all humans as the denominator, he's easily in the top 1% or 0.001% (I haven't followed his work closely, but you'd only have to be a good dev to be in top 1% of the global population).

yeah, I think you could get much of the benefits if Iceberg (or ducklake) added support for `.vortex` (e.g. ducklake only supports `.parquet` right now)

$2.3M–$3.7M per year gross revenue:

Putting it together (podcast‑only)

Ads (STM only): 0.85 – $0.95M

Memberships: $1.5–$2.6M/yr

Working estimate (gross, podcast‑only): $2.3M – $3.7M per year for ads + its share of memberships. Mid‑case lands near $2.9M/yr gross; after typical platform/processing fees and less‑than‑perfect ad sell‑through, a net in the low‑to‑mid $2Ms seems plausible.

---

Quick answers

How many listeners: 175K downloads per episode, ~1.4–2.1M monthly downloads, demographics skew U.S., median age ~36.

How much revenue? $2.3M– $3.7M/yr gross from ads + memberships attributable to the STM show

I was curious how much revenue a podcast I listen to makes. The podcast was started by two local comedians from Phoenix, AZ. They had no following when they started and were both in their late 30's. The odds were stacked against them, but they rank pretty high now on the the Apple charts now.

I looked into years ago and couldn't find a satisfying answer, but GPT-5 went off, did an "unreasonable" amount of research, cross referenced sources and provided an incredibly detailed answer and a believable range.

who is it? give them a plug, seems like it works well.

are they locked-in tho? People loved Arc, but they killed it. Doesn't seem the reviews of Dia are all that great.

How does the scraper work? e.g. LinkedIn aggressively blocks scraping and you'd need to be logged in to see most things you'd care about. How do you handle that?

I agree it doesn't make sense. I'd think they could alias their own tools to match Anthropic's, but my guess is they don't want to customize too heavily on any given model.

CC being better than Cursor didn't make sense to me until I realized Anthropic trains[0] it's models to use it's own built-in tools[1].

0 - https://x.com/thisritchie/status/1944038132665454841

1- https://docs.anthropic.com/en/docs/agents-and-tools/tool-use...

an effective "vectorless RAG" is to have an LLM write search queries against the documents. e.g. if you store your documents in postgres, allow the LLM to construct a regex string that will find relevant matches. If you were searching for “Martin Luther King Jr.”, it might write something like:

    SELECT id, body
    FROM docs
    WHERE body ~* E'(?x)                                     -- x = allow whitespace/comments
      (?:\\m(?:dr|rev(?:erend)?)\\.?\\M[\\s.]+)?             -- optional title: Dr., Rev., Reverend
      (                                                      -- name forms
        (?:\\mmartin\\M[\\s.]+(?:\\mluther\\M[\\s.]+)?\\mking\\M)  -- "Martin (Luther)? King"
      | (?:\\mm\\.?\\M[\\s.]+(?:\\ml\\.?\\M[\\s.]+)?\\mking\\M)     -- "M. (L.)? King" / "M L King"
      | (?:\\mmlk\\M)                                       -- "MLK"
      )
      (?:[\\s.,-]*\\m(?:jr|junior)\\M\\.?)*                  -- optional suffix(es): Jr, Jr., Junior
    ';

we[0] have a pretty complex agent[1] running on Pydantic AI. The team is very responsive to bugs / feature requests. If I had to do it over again, I'd pick Pydantic AI again.

0 - https://www.definite.app/

1 - https://pydantic.dev/articles/building-data-team-with-pydant...