Pinned Tweet
The data engineer job title is due for an update.
Not because AI is replacing the role, but because AI is finally revealing what the role was always actually about.
Moving data was never the point. Meaning it is.
Read more: dataengineeringweekly.com/p/…
Slack is one of the few companies where everyone feels proud of their work. Every scaling assumption we make becomes invalid in 6 months—the team, the energy, the culture, unmatched. Once-in-a-lifetime experience.
Latest (and probably penultimate) Building Slack post is up: buildingslack.com/salesforce…
What it's like being eaten by the whale
What makes a catalog truly open?
When every query engine can seamlessly access the metadata it holds, and that metadata can be freely shared across catalogs—without proprietary barriers or vendor lock-in.
Personal project coding - Claude
Office work - ChatGPT for work
Personal assistant - Muse
Personal Finance - Perplexity
Quick idea brainstorming - Gemini
A clear pattern of frontier market segments emerging.
Ananth Packkildurai retweeted
we have received a lot of dms asking: how 2 people, without a research lab, trained hobbylm using roughly 100 - 400× fewer gpu-hours than many models in this space
so here is the short answer: we spent 76 hours training on 4 h200s, bringing the full base run to roughly 304 gpu-hours.
for context:
sarvam-1: ~122,880 gpu-hours
zoho zia 7b: ~153,600 gpu-hours
llama 3.2 1b: 370,000 gpu-hours
llama 3.2 3b: 460,000 gpu-hours
i will share the engineering decisions, failed experiments, what worked, and the full training story over the next few days.
built with @hariharanms95.
We are releasing HobbyLM from @Fuellabsai : a sparse MoE trained from scratch, with 1.037B total parameters and ~305M active per token, on ~100B tokens.
Here's how we built 🤞
In this post, I take a closer look at the modern data stack and ask how ready it is for AI. I share my thoughts on how the stack needs to evolve as AI takes on writing pipelines and answering business questions.
dataengineeringweekly.com/p/…
For all engineering leaders, how are you building craftsmanship & sense of pride within your teams in the era of AI coding?
Counting is a hard problem, but a $20B margin?
OpenAI annualised revenues $20bn less than previously signalled ft.trib.al/67d4MJe
Ananth Packkildurai retweeted
We just made some of the biggest announcements in @ZetaGlobal's history.
Introducing AthenaOS, our enterprise intelligence operating system that starts with what you want to achieve, and Athena MCP, bringing Zeta intelligence into the tools businesses already use. Day one integrations with ChatGPT, Claude, Palantir and Gemini.
We also unveiled our new Athena Inference Model (AIM) being developed with @FireworksAI_HQ using @nvidia Nemotron, and announced early work with partners on a specialized chip through a capital-light approach.e
What a massive kickoff to #ZetaLive26! $zeta
LG 🚀🚀🚀
Read more here: businesswire.com/news/home/2…
Ananth Packkildurai retweeted
Data Loader (n.): The part of your training stack that is too ML-y for the systems folks and too systems-y for the ML folks.
It's the work I love best, and I had a ton of fun building a new one with the team here at @DatologyAI. My three favorite things about Zephon:
At @datologyai we run thousands of data experiments. Two things keep biting us:
- Same run on 16 vs 8 GPUs → different evals
- Every pipeline change → re-materialize the dataset
Introducing Zephon, the data loader we built to process data on the fly, deterministically. 🧵
The diagram shows where the data should go.
The engineer's question shows where the work will land.
New essay, a field guide for data leaders: Thinking Like a Data Architect
dataengineeringweekly.com/p/…
We should fundamentally rethink the data pipeline orchestration engine for the AI era. Both dagster and Airflow got confused; that adds more complexity than it actually solve
Ananth Packkildurai retweeted
The more time I spend working with coding agents, the more convinced I am that they make software engineering even harder
We can do amazing things with them, but unlocking their full potential requires extraordinary discipline and knowledge
Bury your head in the ground at your own risk. I aim to not jump on any hype trains, but since Opus 4.6 and GPT-5.2 + the harnesses it was clear that these things can write code nearly as good as I can in my best language; better in other languages.
But we won't see non-devs push production code for a long time; possibly forever, if you ask me. Software still needs to be built in robust ways, and it's our profession to do this well, and use the new tools we have, which create new and interesting (+tough!) challenges
Why are AI labs flexing their 'model hacking' workarounds? Stop treating structural vulnerabilities as a badge of honor. It’s an engineering failure. Just fix the bug.
Strogatz's Sync is an interesting read - as AI-driven systems interplay, how agent sync works across the industry is an interesting phenomenon to explore.
Ananth Packkildurai retweeted
The clankers have the keyboard. Rather than mourn the transition, we find it exhilarating.
Announcing After Coding, a new Pebblebed podcast with @keithmadams. Episode 1 with @josh_wills is live.
link >>