Leading AI product innovation and AI impact thought leadership @ Jellyfish. Formerly Harvard, MIT Lincoln Labs. I love running; it doesn't always love me back.
Boston, MA
Joined August 2013
- Tweets1.3K
- Following611
- Followers451
- Likes1.2K
It's kind of insane to think that the typical software org isn't willing to trust 20% of their dev capacity to merge code 🤯
Check out my piece for @tessl_io on the gap we see in the @_jellyfish_co data between agentic coding vs "agentic shipping":
tessl.io/blog/there-arent-en…
Talking about all this at @ainativedev Con in Brooklyn, Nov 2–4. Use code NICKA30 for 30% off: aidevcon.io/
Nicholas Arcolano retweeted
Every time a new frontier model comes out, we go through the same rituals: benchmarks get reported, leaderboards get reshuffled, and the social feeds fill up with a heady cocktail of vibes, hot takes, and everything we’ve found the new model can “one-shot.” Anyone watching from the sidelines could be forgiven for thinking that model performance is the single biggest factor driving AI transformation today.
But day to day, how much does the model really matter?
As part of his work at Jellyfish, Head of Research and AI Nicholas Arcolano analyzes and extracts industry insights from our massive collection of software engineering workflow data, which tracks AI coding tool and agent telemetry across 300,000 developers from 1,300 companies. And when you look at what actually happens inside real engineering teams after a new model drops, the picture is a lot more complicated than the launch-day discourse would suggest.
He shared insights during AI Engineer NYC last week. See an overview of "It's Not the Model, It's You" here: dev.to/dailycontext/its-not-…
Nicholas Arcolano retweeted
The plants all around us are filled with bio-electrical signals that can tell us about the plant, its health, and the environment that we share. What if we could listen?
That's why we built Bombadil - our foundation model for decoding plants' electrical language.
Stop arguing about which frontier model is better, it doesn't matter
(Clip from my @aiDotEngineer NYC talk – cc @_jellyfish_co)
Still annoyed with Claude for embarrassing me in front of @dexhorthy at @aiDotEngineer last week.
Same feeling as when you try to show off your well-behaved kids at dinner and they choose to act like insolent little brats instead.
Nicholas Arcolano retweeted
To make rust build times faster, key to rust's viability as an agentic coding language, I have asked Claude to rebuild it in python
Just made it to @aiDotEngineer, first thing I saw was my story in "The Daily Context". Haven't had a story in a print newspaper since high school 😆
Check it out! @_jellyfish_co insights about what what happens when devs change LLM model families.
dev.to/dailycontext/its-not-…
Nicholas Arcolano retweeted
Did you catch all the live product demos showcasing our new releases during Day 2 of AI Impact Week?
Sessions included:
- "Meet the new Jellyfish, built for the AI-DLC" with Adam Ferrari
- "Measure AI Impact: Where do I stand?" with Tricia Gregg and AJ St. Aubin
- "Transform with AI-DLC: Am I transforming?" with Ryan Servais and Greta Heissenberger
- "Prove AI ROI: What is it worth?" with Taylor Morrel and Noah Flowers
- "How the best engineering teams win with AI" with Ryan Kuchova and Emily Dunn
In case you missed them, we have recapped them all here with videos for you to watch at your own time: jellyfish.co/blog/ai-impact-…
#AIImpactWeek #AIImpact #AIEngineering #ProductReleases
Anyone figured out how to get @claudeai (or any other agent) to just know how and when to "walk you through" its outputs?
Like when I'm doing interactive strategic-type stuff and Claude just vomits a wall of text at me. I always need to say "please walk me through this step by step".
I know it's less token-efficient to take multiple turns, but hate having to read pages and pages of output and then spit back pages and pages of thoughts and corrections – I'd rather do it iteratively but can't get Claude to play ball without asking nicely every time 😥
Biggest problem with Claude is that putting ! in front of terminal commands has become muscle memory and now I can't use a normal terminal
🚨 September cut of @_jellyfish_co AI Engineering Trends is out!
It's crazy, 6 months ago the biggest divide was whether agents wrote your code or not, now the biggest divide is whether you trust them to merge code on their own.
jellyfish.co/ai-engineering-…
It's utterly absurd how far we've come 🤯
Big model release today - I wrote about Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna - plus comparison grids of pelicans by the different model families at different reasoning levels simonwillison.net/2026/Sep/2…
Heading back to NYC *again* in November! Accepted first wave to speak at @tessl_io @ainativedev con
Come join me and I'll give you some sweet @_jellyfish_co data-driven tips on how to stop botsitting. Or maybe how to just babysit way more bots. IDK
Nicholas Arcolano retweeted
🔔 Just Announced: Nathen Harvey (DORA) & Laura Tacho (AWS) to host a Fireside Chat at AI Impact Week
Join them on Day 1 as they discuss the trends in how AI adoption, assistance, and agents are paying off across your teams. You'll hear them discuss topics like:
- What's the AI-adoption problem engineering leaders aren't admitting to yet?
- As AI takes over more of the day-to-day coding work, where do you expect bottlenecks to show up in 2027?
- Which metrics are quietly lying to engineering leaders right now, and what should replace or supplement them?
- As engineers grapple with the changing dynamics of their job, what can leaders do to keep them engaged and excited about the work?
- What's the one thing an engineering leader should measure or change this quarter, not next year?
Don't miss the chance to hear from these developer experience luminaries...register today: jellyfish.co/events-and-webi…
#AIImpactWeek #AITransformation
Awesome episode @swyx – Jev is pretty nuts and @CompleteSkeptic even moreso 🙂
Considered heading back home when I got to work so I could do my commute twice and keep listening to it
Jev and the System One Model: RLCD, intelligence/$, reliable AI, & the end of chat-first AI latent.space/p/jev
@typesafeai CEO @CompleteSkeptic explains why AI can solve extraordinarily hard problems yet still fail to automate basic work, why Jev is built for reliable decisions inside software instead of chat, why TypeSafe rejects public benchmarks and refusals at the API layer, why data and the right task matter more than brute-force compute, how System One Models could reshape coding agents and software, and why even with $1 billion he wouldn’t pre-train a model from scratch.
Getting psyched for @aiDotEngineer in NYC next month!
Everyone's talking about how AI might end civilization. I'm going to talk about how in the real world the model you pick doesn't matter very much 😉
wish I was going to @aiDotEngineer Paris instead but hey NYC is pretty great in the fall too