Matthew Mullins
banner
mmullins.coginiti.co
Matthew Mullins
@mmullins.coginiti.co
technologist, philosopher, and outdoors enthusiast

CTO@COGINITI
This is the answer 👆! One Drive is awful
February 12, 2026 at 7:54 PM
Trigger runs on schedules or ticket events, automate Slack notifications, and keep an auditable trail in Jira all while your actual data logic stays governed and reusable in the Coginiti catalog.
February 12, 2026 at 2:16 PM
We just shipped a verified Coginiti node for n8n. That means you can orchestrate Coginiti catalog assets and run SQL/CoginitiScript inside n8n workflows, right alongside the systems teams actually live in like Jira, Slack, and Snowflake.
Announcing the Coginiti node for n8n
he Coginiti node for n8n makes it easy to orchestrate governed analytics workflows—run scripts and blocks, automate notifications, and keep work standardized.
www.coginiti.co
February 12, 2026 at 2:16 PM
The views don't suck
February 11, 2026 at 12:44 PM
I don’t know how many people even knew there was litigation over the name. I dont know how it was resolved
February 9, 2026 at 11:59 PM
Yes, any wonder if they lost their lawsuit with coalesce.io
February 9, 2026 at 10:12 PM
Thanks Paul
February 8, 2026 at 3:13 PM
I continue to be unimpressed with the open semantic interchange (OSI) so I send messages into the void. #databs
How Not to Run an Open Standards Initiative
A clear-eyed look at how open standards fail: vendor capture, weak governance, and specs without institutions to support real interoperability.
open.substack.com
February 8, 2026 at 2:56 PM
I love your eloquence Paul!
February 7, 2026 at 3:36 AM
Don't know what Kalshi is and I'm afraid to ask
February 7, 2026 at 2:59 AM
Hey #databs help me out with some data collection. We've got a little survey going and I'd like to fill it up with feedback from folks that aren't already customers. Takes 5min and I'll pay you back in gratitude and like all your posts. www.surveymonkey.com/r/FSVBR9Z
February 6, 2026 at 6:25 PM
Reposted by Matthew Mullins
Men should read Absolution apparently. Men men men men men. Whatever. www.insidehook.com/books/read-r...
What to Read Right Now, According to Cool Men
We talked to journalists, authors and editors about books men should read
www.insidehook.com
February 6, 2026 at 2:54 PM
Reposted by Matthew Mullins
Anthropic to developers: Claude Code makes you more productive when building SaaS apps.

Anthropic to businesses: Our AI agents make SaaS apps obsolete.
February 6, 2026 at 3:00 PM
Being an ASF project rather than an LF project puts some stronger guardrails how DBRKs can steer the project. There is a pretty broad developer group even post acquisition, and it appears DBRKs strategy is to move the best features of Delta to Iceberg.
February 4, 2026 at 11:00 AM
Lot of chatter about AI ready data and how the shape of data might be changing. It's not, but your data management skills need to be on steroids. #databs
AI-Ready Data Is Still Analytics-Ready Data
AI-ready data isn’t new—it’s analytics-ready data done right. Learn why clean models, semantics, context, and SQL pipelines matter for AI systems.
www.coginiti.co
February 3, 2026 at 1:27 PM
It’s really not though
February 2, 2026 at 12:28 PM
I was looking for something to watch, but not sure I'm emotionally prepared tonight
February 1, 2026 at 2:59 AM
That seems to support what Thompson is saying in her piece. The agents aren't going to give you the architecture and abstractions on their own. At least not good ones.
January 29, 2026 at 5:03 PM
The way I said that may have sounded too combative, I'm not looking for surrender but to see what actually works.

If you're providing the agent with a TLA+ model as a pattern for it to follow, that's a leg up. A jr isn't likely to know what TLA+ modeling is, how to use it, or how to compare models
January 29, 2026 at 5:03 PM
I'm not saying you're lying, but bring the receipts if you've got them.

I've had good success on small greenfield projects, but that's it. My experience is mirrored by most of the peers I've talked to.
January 29, 2026 at 2:35 PM
It can do replication, but it doesn't do good generation out of the box. If you're a jr who doesn't have a good grip on architecture, the model won't do the work for you.

Admittedly a lot of every day devs aren't good at these things either, but we have architects and staff+ for a reason.
January 29, 2026 at 2:10 PM
It wouldn't be weird if everyone was telling you that bandsaw's can make chez lounges along with hammering all the nails in your house.

I think agents can be useful tools, but they certainly come with some serious limits. They aren't a replacement for the engineering aspect of our discipline
January 29, 2026 at 1:30 PM
Say it loud… “AI coding agents can produce syntactically correct code. However, they don’t produce useful layers of abstraction nor meaningful modularization. They don’t value conciseness or improving organization in a large code base. We have automated coding, but not software engineering.”
Breaking the Spell of Vibe Coding – fast.ai
Sinister variations on the positive state of flow
www.fast.ai
January 29, 2026 at 12:36 PM
I mentioned unconferences to someone this weekend and they thought I was making it up. So, yes we should bring those back
January 27, 2026 at 10:02 PM
I said this at the beginning of the year. Bring The Receipts! I want real numbers, not vibez
“[O]ne said the platform has made him five times more productive.”

Going to need numbers here, not vibes. I don’t care — at all — about developer feelings, about their perceptions.

If this is such a ground-shaking development where are the illustrative studies? Show me numbers.
Anthropic’s Claude Code Platform Looks Like a Software Killer
Claude Code promises to create any type of software — from analyzing specific data sets to building websites and editing photos.
finance.yahoo.com
January 27, 2026 at 8:32 PM