@igor
Account based inCanada
About this account
- Account based in
- Canada
- Connected via
- Canada App Store
Account-level information from X, not a live location or the device used for a specific post.
Pinned Tweet
Fastest commerce marketing teams are using AI to move at unprecedented speed. Check out the latest @SuperpilotAI case study with ALICE + OLIVIA's NYFW campaign 👉 superpilot.com/customers/ali…
Igor Faletski retweeted
just got off the phone with a new engineer hire at a portco with a very wise insight
“AI has been replacing my job since I got a CS degree, yet I’m busier than ever - we can all just be more ambitious”
this is the same insight that the inimitable @JensenHuang has for us all
Igor Faletski retweeted
People rest when things are settled. Nothing about this is settled. Every week the map redraws itself and you either keep up or wake up irrelevant. That's exhausting and also the most alive I've felt in this industry. Both things are true at once.
In my first company, as a founder in early 20s with compsci background, I actively seeked out experts in GTM, sales, FP&A just to understand how those functions work.
Today I can get any specific question on those fields answered at a very high level of proficiency in a few keystrokes with Claude on Saturday morning.
What does it mean for human start-up exec expertise? It’s very important to be extremely current about where the edge in your field has been in the most recent 3-6 month window, and knowing how to practice it. Having a human network to bring in talented hires. Ability to execute, not just pattern match.
The opportunity is only bigger for the best of the best - and not there the way it used to be for those underinvesting in staying *very* current.
Igor Faletski retweeted
the potential of the current moment makes me feel a little physically ill
maybe it’s lack of sleep
Salesforce’s excellent results are a timely reminder that having years and years of system of record data *is* an AI moat after all. Having Slack doesn’t hurt either!
A less obvious fitness heuristic: your ideal exercise schedule is a function of your perfect sleep schedule.
Lots of buzz this week about ChatGPT downgrading the share of Reddit content in its answers compared to brands' owned and operated sites. It really puts the focus back on web content production pipeline and raises the bar for what retailers in particular need to be doing. 1-2 pages launched per month is not enough.
In our experience, best-performing content marketing teams are not held up by legacy systems and processes:
- Content creation is agent-first, with human users providing creative vision and oversight.
- The whole workflow is automated with agents end-to-end (not just text generation alone): if multiple teams are involved in content production and approval, the AI has multiplayer UX in place.
- Performance and benchmarking data informs every step of campaign prioritization, publishing and ongoing optimization, vs sitting in a silo'ed dashboard somewhere
- All of this is operating on an autonomous loop directly tied to traffic and GMV trends, pulling in the marketers proactively to move much faster overall.
If you're interested in adopting a modern agentic marketing platform to get ahead with AEO/GEO/SEO and grow GMV faster, send me a dm!
For marketing agents to really excel at commerce content, they need to master understanding and selecting brand images. To unlock this, we just shipped @SuperpilotAI's Agentic Image Search. Check out the latest blog post: superpilot.com/blog/how-to-a…
Igor Faletski retweeted
It makes sense to optimize model routing at the harness layer instead of the gateway layer if you want to hillclimb on accuracy/cost for any e2e task.
Every task is solved by a combination of a model mixture and agent harness. Every task requires a different mixture of models (+harness logic) to be at the pareto frontier of accuracy and cost.
* If you only optimize the model mixture at the gateway layer, you lose the broader context encoded in the harness and only optimize at the LLM completion layer.
* If you optimize the model mixture at the harness layer, you can make the "optimal" model choices a priori and while in the agent loop
For any given task, the model and harness are probably co-optimized together, so that the model mixture can only exist with this specific harness shape and vice versa
Leaning too much into the human employee analogy (“hire my agent!”) for AI companies is dangerous - there are simply too many structural differences that can’t be masked with messaging.
Much better to take the emerging AI messaging conventions and customize them even better for your target vertical - terminology, outcomes, et al.
If there’s a universal piece of advice, it’s “think for yourself”. The time-space continuum warps and the simulation is uneven. You are in charge of you, and what works elsewhere is a mere data point.
💪 Go @dennispilarinos, go Vancouver, go SFU, all of it! 🚀
Vancouver has become one of the best places in the world to build technology companies. Thanks for the opportunity to reflect on the journey from Microsoft and AWS to buddybuild and Unblocked, and why I still believe Vancouver’s best years are ahead.
techcouver.com/2026/07/22/de…
