Dr Paul Kruszewski

According to @RayDalio 's corresponding test, I am a Shaper. The assessment tools really helped me to better understand myself.

Wow, it's quite an achievement for Canada to have had even lower growth than Germany during the last decade. Germany has done almost everything humanly possible to ruin its economy. How has Canada been able to beat that?

@GregEstes @MayfieldFund @NavinChaddha well done @GregEstes

Very excited to support @MayfieldFund @NavinChaddha in growing the AI ecosystem. One thing I love about the Mayfield team is the support they give to their Founders at every stage of their business. It's always about the people, and they are great people. linkedin.com/feed/update/ur…

what an amazing turn around! More please indeed!

You know taxes are high When we are beating France 👻

Fantastic work by Dr. @jackmintz & the @CDHoweInstitute: Canada needs big-bang tax reform We should cut top income tax rates and let people opt for a generous credit that pays them more than applying for boutique exemptions financialpost.com/opinion/canada…

@StephenKing Me too Am I that old?

Have used every fucking one.

Remi and team are the real deal Proud to support them

Brilliant Harvest (@B_HarvestInc), the Calgary firm that runs an AI-driven platform for the heavy equipment industry, describes itself as “farmers first and innovators second.” It just secured $4M USD in seed funding. Here's what it plans to do with it: betakit.com/agtech-startup…

Whatever happened to the extremely cool 80s TV genre of "Shows based on a single extremely technically advanced vehicle"? Airwolf, Knight Rider, Viper, Blue Thunder, etc. It was an amazing genre that should return to inspire modern youth. youtube.com/watch?v=9Jbbxz…

@conor_ai @jia_seed Minor character; like main character but with less zeros

@jia_seed Tizz/Rizz explainer (filmed on uber ride)

the tizz / rizz founder matrix (TRFM) all great founders land somewhere on here tag yourself cooked up with @jia_seed

Sad but true

@goclio continues to set the bar for #cdntech

Clio (@goclio) to acquire American legaltech @vlex for $1 billion USD. #CDNtech #BCtech betakit.com/clio-to-acquir…

@danperez610 and the @HingeHealth team rock!

We're checking up on @HingeHealth 🩺 CEO @danperez610 shares how he's aiming to transform the physical therapy landscape as the company makes its market debut at the @NYSE: cnbc.com/video/2025/05/…

@karpathy makes total sense

We're missing (at least one) major paradigm for LLM learning. Not sure what to call it, possibly it has a name - system prompt learning? Pretraining is for knowledge. Finetuning (SL/RL) is for habitual behavior. Both of these involve a change in parameters but a lot of human learning feels more like a change in system prompt. You encounter a problem, figure something out, then "remember" something in fairly explicit terms for the next time. E.g. "It seems when I encounter this and that kind of a problem, I should try this and that kind of an approach/solution". It feels more like taking notes for yourself, i.e. something like the "Memory" feature but not to store per-user random facts, but general/global problem solving knowledge and strategies. LLMs are quite literally like the guy in Memento, except we haven't given them their scratchpad yet. Note that this paradigm is also significantly more powerful and data efficient because a knowledge-guided "review" stage is a significantly higher dimensional feedback channel than a reward scaler. I was prompted to jot down this shower of thoughts after reading through Claude's system prompt, which currently seems to be around 17,000 words, specifying not just basic behavior style/preferences (e.g. refuse various requests related to song lyrics) but also a large amount of general problem solving strategies, e.g.: "If Claude is asked to count words, letters, and characters, it thinks step by step before answering the person. It explicitly counts the words, letters, or characters by assigning a number to each. It only answers the person once it has performed this explicit counting step." This is to help Claude solve 'r' in strawberry etc. Imo this is not the kind of problem solving knowledge that should be baked into weights via Reinforcement Learning, or least not immediately/exclusively. And it certainly shouldn't come from human engineers writing system prompts by hand. It should come from System Prompt learning, which resembles RL in the setup, with the exception of the learning algorithm (edits vs gradient descent). A large section of the LLM system prompt could be written via system prompt learning, it would look a bit like the LLM writing a book for itself on how to solve problems. If this works it would be a new/powerful learning paradigm. With a lot of details left to figure out (how do the edits work? can/should you learn the edit system? how do you gradually move knowledge from the explicit system text to habitual weights, as humans seem to do? etc.).

@gregestes It was a pleasure to work with you

Well, I've just retired from NVIDIA after 15 years. Thank you to all my amazing friends, colleagues and partners over 40 years. What a great company, filled with brilliant people doing their life's work. Humbled to have been part of it, and will be cheering them on every day. linkedin.com/posts/estesgre…

Excellent editorial

Globe editorial: The prime minister that Canadians need /via @globeandmail theglobeandmail.com/opinion/editor…