Derek Manuge
960 posts
Derek Manuge
@Manuge
3x Fintech Founder. 2 Exits. Investor. Operator. Engineer. Rational Optimist. 🇨🇦 🇺🇸 🇰🇾
Toronto, Ontario Katılım Ekim 2012
2.5K Takip Edilen518 Takipçiler
Cue the “Average SP500 Returns After $VIX > 30” charts. Here’s ours: muhat.com/blog/volatilit…
English
I’d like to see the source of that statistic. It is claimed in the article as analysis - but whose and where? I can’t find it.
CutTheCrapInvesting@67Dodge
Wow, ouch. Oh Canada ...
English
@ddebow I'm honestly confused on why we aren't keeping repeat offenders in jail. Why would a competent government not want this if they truly cared about the well-being of their citizenry? And yet our government actively votes against these proposed motions.
English
45 prior charges?!
come on people, what are we doing here?
Rob Khazzam@rkhazzam
Canada doesn't need to be this way. Violent home invasions are ruining what semblance of peace and safety we previously had. If you've been charged 45 times including for trafficking cocaine and fentanyl, multiple violent armed robberies and charges for MURDER you should be in jail. Bail should simply not be possible. Law abiding Canadians "didn't sign up for this" and if it continues we will see more people leave and more vigilante justice. We have to choose rationality, not stupidity.
English
Another trait I noticed about 4.20 is that it can suffer from "over confidence" with answers (which Gemini is notorious for). When several answers are simultaneously true, it can sometimes "go all in" on one. Perhaps it's preference, but if we are optimizing for truth, a more balanced response or a trigger to obtain more information from the user might guide towards resolving the uncertainty in choice. I see this as a post-training calibration issue so I'm sure it's easily solvable.
English
@elonmusk Honest feedback as a Grok power user...Expert 4 is better than Expert 4.20 for thinking and reasoning (more so with financial and legal matters). I would prefer to use version 4 if I had the option. @forwarddeploy
English
What are your initial impressions of Grok 4.20?
Major upgrades are still landing every week.
Testlabor@testerlabor
Grok 4.20 is now officially out of Beta. It's now on Auto, Fast, Expert & Heavy.
English
@danielfoch Lots of factors, but it's environment and incentives. Canada does not have the same risk-taking culture as the US (just look at the LTGE vs. QSBS limits). My advice to founders is that if you want to raise capital or sell, just go to the US.
English
Interesting. Why so few Unicorns? 🦄
FounderEric@FounderEric
The world's number of startup accelerators 1. 🇨🇦 Toronto - 77 2. 🇺🇸 New York - 63 3. 🇺🇸 SF - 62 4. 🇬🇧 London - 44 5. 🇨🇦 Montreal - 44 6. 🇩🇪 Berlin - 37 3. 🇺🇸 Boston - 23 7. 🇨🇦 Waterloo - 21 8. 🇨🇦 Vancouver - 28 9. 🇺🇸 Miami - 19 10. 🇺🇸 Austin - 17 fundingcake.com/startup-direct…
English
English
Shopify Builder Sundays in Toronto 🍁
Coffee, snacks, and great views from the @Shopify office
English
Derek Manuge retweetledi
“A people that has forgotten its traditions, its ancestors, and the sacrifices that bought its freedom is a people already conquered. For the man without memory is easily ruled, and the man without roots is easily led.”
—Hilaire Belloc
English
100%. I felt the same way again when DeepSeek (et al) "introduced" reinforcement learning in their models a few years ago. I remember building trading algos 15 years ago based on John Moody's research (pioneer of Recurrent RL) and assumed RL (or a variation of it) was being used by the foundation labs. Such an obvious way to improve model performance. Makes me wonder what other techniques are being ignored because each new model is either a marginal iteration of the prior model version and/or a retraining with more compute.
English
I honestly thought the foundation labs were already doing this internally.
We were doing genetic algorithm optimizations for computer simulations in 2004-2008, and the technique was already mature.
I wonder how many other techniques are hiding in plain sight.
Andrej Karpathy@karpathy
I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autor… Part code, part sci-fi, and a pinch of psychosis :)
English
@REDWRITER @Dogetothemoon Do you have any details on how the X card handles FX (if at all)?
English
Glad to see a fellow shibe, @Dogetothemoon, join the X Money pack. Dogecoin + X Money feels right
English
The key to X Money's success will be if they can offer a truly "global" card with no FX fees and near-market exchange rates. The problem with many of the market-leading cards today is that they are only open to US residents (or those with a valid TIN) and the non-US fees make them unusable when travelling... I'm sure this is the focus of @nikitabier @elonmusk . Excited for it to happen.
Josh Kale@JoshKale
X Money just went live in beta and it’s gotta be the single best financial product on the market today These numbers don't even make sense: - 6% APY on deposits - Insured up to $250,000 = $15,000/yr in interest, $1,250/mo when maxed - The metal card gives 3% cashback on all purchases matching Robinhood gold Also equipped with direct deposit support from your day job and creator payout support from your posting Everything flows into one account. Earning 6% APY on the way in, and 3% on the way out. I'm not sure how sustainable this is but man is it compelling. images from @REDWRITER who's been sharing some great stuff
English
@tylermeredith "For standards of living it is a poor insight. Since petro states and tax havens do exceedingly well on it." ...a simple search would show that countries like UAE, Bermuda, Monaco all have exceptionally high standards of living, exceeding that of Canada.
English
No one says GDP or growth doesn’t matter. Thats ludicrous.
But since for several years this one stat has been abused to say a lot of things - it’s pretty darn important to know what it does and doesn’t tell us.
For standards of living it is a poor insight. Since petro states and tax havens do exceedingly well on it.
This is why things like quality of life reporting - something the Government of Canada has been experimenting with in policy application for several years (ask Lucy) - is important.
If we want to fix: productivity, inequality, and health outcomes — we need to properly focus on those.
Canada is a high wealth / high income society with fairly good health outcomes. That has challenges with productivity, and is increasingly high cost for basic needs.
GDP per capita is a poor insight into those issues and what to do about them.
#cdnpoli
daniel debow@ddebow
this whole gdp/captia / alabama argument seems ... weird. is the @tylermeredith @JimboStanford argument that GDP is meaningless? Or that everything is a-ok in 🇨🇦? sure gdp/capita - like ALL measures - has limitations. but its not useless. its a powerful measure of prosperity, but not the only one. but if you consider the *trend* in other measures, as @MikePMoffatt suggests, and more importantly - if you open your eyes and ears to the state of our cities, our young people, our entrepreneurs - one can see we're not getting better fast enough to fund and improve the "other measures" which are mostly delivered through redistributive social programs. so the point of Alabama (or any) comparisons is to consider: what can we learn? what are good ideas we can adopt? what can we do to accelerate growth? its not to be Alabama of course! its to be Canada - and get much more prosperous, faster.
English
Possibly via Collosus' network? Otherwise the solution needs to work within the Visa/MC network and that is the source of the fees. Ultimately the merchant wants local currency (MXN) vs. local stables, which is the challenge for almost all markets (except e.g. USD=USDC, EUR=EURC).
English
English
We pay for the privilege to increase our dependency on others until there is nothing left to pay.
Mike Solana@micsolana
this all becomes easier to understand once you realize we're not paying taxes for public infrastructure or services, we're paying taxes to increase the number of bureaucrats who vote for the democratic party
English
Game theory will become more important than ever. Humans over index on trust, but AI won't. In other words, "reputation" will become exploitable in the early stages of AI-to-AI commerce. Even if formal reputation systems exist, commerce flows must be designed to handle trustlessness throughout the buying journey to minimize the attack surface -negotiating terms (e.g. prices), facilitating payments, delivering goods, managing returns and disputes, etc.
Beff (e/acc)@beffjezos
The AI-to-AI economy will far outpace the AI-to-human, human-to-AI, and human-to-human economy in short order
English
The optimal genius-to-idiot ratio is different for everyone, but if you're not feeling like an idiot enough, you're not pursuing audacious-enough goals.
Dream bigger. Be idiotmaxxing.
Lior Alexander@LiorOnAI
Just had coffee with a founder who sold his startup for $300M. He confirmed what I always felt: To achieve anything, you’ll swing between “I’m a genius” and “I’m an idiot.” Sometimes weekly. Sometimes hourly.
English
If you're a business owner and not using LLMs to verify the accuracy of your accountants and lawyers, you're NGMI.
Ramp Capital@RampCapitalLLC
Has anyone fed their previous years’ tax returns + docs into an LLM and used the output to help fill out this year’s taxes?
English
Derek Manuge retweetledi
