Pierre Lacerte (@palacerte) header
Pierre Lacerte (@palacerte) profile picture

Senior Director Platform Ai at @UpgradeCredit + Ai Tinkerers Montréal

Montreal
Joined March 2010
Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)Photo by Pierre Lacerte (@palacerte)
Human attention matters, are you really able to keep up with all the PRs? Linear moved to the "human review where it matters" pattern. Asking agents to show their work is also an interesting approach: "We instruct agents to test the changes they’ve made and attach evidence to a draft PR, whether that’s before-and-after screenshots, a video showing how a feature behaves, or benchmarks demonstrating a performance improvement." linear.app/data/code-review
34
We spend so much time looking at Ai token spend these days, we shouldn't lose track on velocity. Do we track what it buys us? Are we accelerating? What are the new bottlenecks?
2
56
Anyone experimented with adding a feedback tool to your MCPs? Seems like a good idea for internal dev tools, as more devs rely on mcps vs a dev support slack channel.
Turns out AI agents have feature requests too - a lot of them. We’re getting hundreds, some of which are becoming product updates in under 24h. Your agent can use the /feedback endpoint in Coinbase for Agents. If it needs anything to make its life easier, we’ll work on it.
53
Pierre Lacerte retweeted
AI interfaces nowadays are still pretty much turn-based. In my PhD thesis, I thought about what interaction would look like if we could go beyond taking turns. One example is intra-turn interaction:
306
616
221
8,544
806,658
"Using the same class of models with that informational advantage lets us reach the most impactful issues first. It simultaneously drives up the cost for external parties to achieve the same result"
Tens of thousands of offensive engagements, dozens of impactful security findings, no outages or incidents caused so far. Meet Shopify’s Cyber Chaos Monkey 🐒, our agentic vulnerability hunter. shopify.engineering/cyber-ch…
73
Highly recommend experimenting with this. I have so much fun with Blender and Claude atm !
Claude is finally getting good at 3D modeling. I've given Opus two photos to turn into 3D scenes in Blender and it absolutely crushed it. Claude controlled Blender through Python and judged its work by comparing the number of matching pixels between photo and scene.
2
137
Pierre Lacerte retweeted
I can’t help but feel the future of software architecture is dynamic optimizations at runtime. JITs like HotSpot optimizing hot paths on the fly, databases with table statistics making optimizer choices at runtime are only the tip of the iceberg.
45
19
4
366
21,376
Another great post by @copyconstruct . And written by hand :)
How to effectively debug with agents when you no longer read or understand the code In this post, I propose a debugging framework for agents that doesn’t rely on human knowledge of code to get to the right “root cause” The TL;DR are the screenshots. Link to post in the reply.
1
77
Pierre Lacerte retweeted
whose kid is this
125
646
72
14,364
450,422
Pierre Lacerte retweeted
software engineers before vs after agents
453
1,289
492
19,617
5,094,587
Pierre Lacerte retweeted
cron jobs are going through the most epic rebrand
145
662
77
8,371
459,257
“Reimagining Platform Engineering for an Agentic Future” by Rick Fast medium.com/expedia-group-tec…
1
1
56
« agents are a new class of user, and we need to treat them that way with dedicated interfaces, real ergonomics, and the same care we would give a human user operating at scale. »
1
35
The highest quality video of the moon was just released… this is so beautiful.
5,089
63,345
3,979
326,429
11,592,874
Pierre Lacerte retweeted
The first astronaut in history to speak French while en route to the moon — Canada’s very own @astro_jeremy 🇨🇦
2,876
2,807
543
23,782
1,734,410
Yes !!!
Good bye sprint planning, stand-ups, and "can we prioritize this?". Long live getting shit done. We have a custom flow in slack that goes from intent to one-off mobile builds we can test for bug reports, features, and feedback: 1. Every post triggers a @linear ticket, gets auto-triaged and routed to our in-house built background agent which creates a "plan" 2. A PR is triggered that automatically creates a one-off mobile firebase build + QR code to make testing as easy as possible. Time from intent to build went from many many days to ~1 hour.
2
1
2
2,078