Host @Boazashkenazy explores AI's impact on work through interviews with AI thought leaders.
Seattle, Washington
Joined February 2025
- Tweets29
- Following1
- Followers7
- Likes6
Most people still talk about AI coding tools like there's a model race to be won.
GitHub's own numbers say otherwise. @mariorod1 Chief Product Officer at @github shared that paying Copilot customers use somewhere between 27 and 29 distinct models, and not one of them accounts for more than a quarter of requests.
That means the fight isn't for the best model. It's for whoever builds the system that routes work to the right one at the right moment.
Full conversation at
shiftai.fm/episodes/114-buil…
Most companies believe they have to clean up their data before they can seriously adopt AI.
@EmrecanDogan5, Head of Product at @glean calls that a pipe dream that will never finish.
I recently had a chance to sit down with him and his pitch to CIOs is the opposite. Leave the messy data exactly where it is. Figure out what's important, what's outdated, and what contradicts itself, and build the index on top of that.
That's a much harder engineering problem than a cleanup project, and it's why Glean spent years working this out before AI made it urgent.
Full conversation on ShiftAI.fm or whever you listen to your podcasts
Link in comments.
Seattle is home to one of the more interesting physical AI companies working right now: @carbon_robotics led by CEO @paulmikesell
Its laser weeder can be shown a weed it's never seen before and just get it. No retraining, no new dataset. A farmer can point at a weed onscreen in the field, say kill this, and it does. First try.
Most vision models still need to have seen a thing before they recognize it. This one generalizes like it actually understands what a plant is, not just what a labeled photo of one looks like.
A laser doesn't care what the weed is, it just needs to find it. Get that right at scale and you're not spraying a field to kill one plant in ten. Fewer pesticides in the ground and in our food, moves from a marketing line to a side effect of the tech actually working.
Full conversation on ShiftAI.fm
Link in comments.
@praerit helped build Active Directory at Microsoft and was the former CTO of Smartsheet. Now, as CEO of @OneIdentity, he's making the case that AI agents don't need a new identity system at all.
His argument is that agent identity is really no different than a human's, it just moves much faster. The problem isn't inventing new infrastructure for agents. It's applying the identity discipline companies already have and never quite finished.
That's a smaller ask than most of the industry is pitching. It might also be the more honest one.
Full conversation on ShiftAI.fm.
Link in comments.
Full episode with PG can be found here - youtu.be/2_YLECXsetE
Big thanks to @BakerTillyUS @wilsonsonsini @Stifel for sponsoring and supporting the Seattle community during Seattle Tech Week 2026.
Most security conversations about AI agents still focus on the perimeter.
@sabina_anja Chief Technologist at @VMware by @Broadcom thinks that's the wrong place to look.
Assistants wait for you to approve every step. Agents don't. Once you set one loose, in her words, "it's loose," and only then do you have to build the guardrails back.
That flips a security model that was built to react, not to anticipate intent.
Full conversation on ShiftAI.fm.
Link in comments.
Shift AI retweeted
AI agents are getting real permissions inside enterprise systems. Identity just became the control layer for the entire agentic era.
Live @shiftaifm recording Thursday July 30 with Praerit Garg (CEO, One Identity, ex-CTO Smartsheet). Seats limited.
RSVP: luma.com/p848dn4z
Shift AI retweeted
I had an amazing time yesterday watching the live recording of the @shiftaifm podcast hosted by @boazashkenazy in downtown Seattle!
No human on call. No human deploying. No human debugging at 3am.
@Hyphen CEO @jaredwray joins @boazashkenazy to make the case for autonomous cloud, AI agents running production infrastructure end to end.
He's not theorizing, he's shipping it.
🎧 ShiftAI.fm
LLMs are confident. Not always right.
@logic_int CSO/COO @PRHillmann joins @BoazAshkenazy on Shift AI.
Sudoku test, leading LLMs vs. Logical Intelligence's model:
LLMs: 2% solved / $14,000
LI model: 98% solved / $4
ShiftAI.fm
Enterprise AI doesn't fail because of bad tech. It fails because the foundation wasn't ready.
Cynthia Tee, former CTO of Smartsheet, on governance, data transparency, and why bold beats cautious right now.
Full episode at ShiftAI.fm
What if your city government actually spoke your language? San Jose built it. Full conversation on ShiftAI.fm
🎙️ NEW SHIFT AI EPISODE IS LIVE NOW!
Round two with Varun Puri, CEO of Yoodli AI Roleplays! The Madrona-backed startup continues to evolve, and Varun shares their completely refreshed go-to-market approach.
Listen/watch now at - youtu.be/qqQblw0UzY0
Excited to be interviewing @davidshim today at @HumanXCo A lot has happened since we last caught up so I’m excited to hear his thoughts and update.
Last night the @humanxco opening ceremony kicked off with @KamalaHarris speaking about the importance of responsible AI and government oversight.
Host @boazashkenazy is down in Las Vegas at the @HumanXCo interviewing AI leaders in the press room. Stay tuned for more updates.
AI prompting is changing Hollywood:
"Productions will use this more and more...you can produce final looking images with prompting. It will change how directors work," Matthias Wittmann, VFX Director at Digital Domain.
Full interview on YouTube Shift AI Podcast EP 53!
In EP 53, Matthias Wittmann (@DigitalDomainDD) predicts exponential growth in media due to AI innovation.
Watch the full episode on Youtube and Spotify!
#ShiftAIPodcast #AI #VFX