data-ottawa

LLM reinforcement learning improves those tasks by having the agents "reason" until they hit a verifiable output state that passes a black or white test. So they learn language which is tuned to the task in a silo, never communicating with anyone or anything.

In contrast, most human communication and writing is intended for other humans, and we constantly hit failures here that we learn to iterate on and improve. Humans who do math do so with the constraints that their math needs to be reviewed by another human and in the context that all their writing needs to be understood by another human.

The process causes a warped and compressed version of language.

Yes, but to be pedantic, no, this is a short term rise in CVEs.

This is a shift in the equilibrium of difficulty to find these exploits, but there are increased defensive tools and new awareness of these issues as important. There will be a large glut of AI detected bugs, there are decades of code surface to scan and clean up, so this will last a while.

Eventually net new software will end up being written in a more secure way, and important old software will be patched or replaced, or taken offline.

qwen-3.8-flash-next is suprisingly good, I find it overthinks, but it's quite usable.

In Vibe CLI app, when you load Mistral Large 4 it swaps out the cat in the header with a larger one

It's a nice little detail

I updated my vibe config to point to Mistral 4 and it charges my subscription capacity, not my API allowance. So it is included in subscriptions, you just need to do a little bit of legwork.

C’est ne pas une ~~pipe~~ Mona Lisa

https://en.wikipedia.org/wiki/The_Treachery_of_Images

The lawyers I’ve used have always asked and used encrypted email. I don’t know if that’s regional, but it was taken very seriously here.

I’ve read this comment four times over and I have no idea what your point is.

Canada has a massive amount of untapped potential, and it’s time we tap it.

How is the primary way of interacting with your computer being good a gimmick?

I try to use my partner’s windows laptop and it’s borderline unusable for me. She only almost uses it with a mouse plugged in. Meanwhile I don’t think I’ve ever seen someone plug in a mouse to a MacBook that wasn’t docked to a monitor and keyboard.

Similar experience, the first thing it did for me was update production data models without a confirmation. It has always shown me the command is going to run and then I confirm.

When most people have a bad experience at a restaurant they never go back, so the harm is hard to quantify until it stacks up.

This was my thinking. It seemed easier to find the exact instructions for my dishwasher and most of the advice available was for newer models that had different control layouts.

Plus, the repair manual is full of flowcharts for diagnosing and resolving issues, which is very useful.

So two regions declare they want to be closer and your reaction is “we should conquer a couple countries just to show them up”?

This is exactly the problem.

What specifically do you mean “under the wings of the US?”

So the implication is continue doing that?

I’ve always been told don’t let perfect be the enemy of good, and the journey of 1000 miles starts with a single step.

The lesson here is that you can’t rely on the US anymore. Not for military protection, not for defence procurement, and not for trade or supply chain. That legacy has ended entirely by choice of the United States.

The problem has moved beyond monetary trade value and into existential risk.

A few weeks ago my dishwasher broke, flooded the kitchen, and refused to drain. I went to the manufacturer’s website to download the manual so I could drain the basin, except then Cloudflare flagged me as a bot for suspicious activity trying to download the right PDF (there’s the dimensions, spec sheet, installer manual, and user manual) and I couldn’t access the manufacturer’s website anymore.

Ironically, I could get the right manual through an aggregator that scraped appliance manuals and slapped ads on the page.

To me that perfectly encapsulated the issue. Clearly the manufacturer wants owners to have access to manuals, that’s why they have the portal. But due to bots, they pay an intermediary to hold their site ransom from users. And then the bots still clearly can access it anyways.

> In fact, with Astra and Fable I feel like not only are the costs astronomical, but the models are also just not for me as a software engineer.

This is how I’m feeling as well. The models are so happy to run and implement that they forget to keep me in the loop or make it understandable.

If I want to magic something into existence or one shot an analysis, the model is happy. But if I want to direct, understand, and manage they fall completely flat and are unable to explain things for me.

It’s even hard to discuss planning because Claude will run off and start pulling data and analyzing my projects and spend 5 minutes on something not needed, when I need a rubber duck that can sanity check things and push back.

My non technical colleagues love it, but I’m fighting the thing all the time.

It’s a contrived example, but also not very far from the mark.

The spinner with random claudisms is what makes the game, mixed with Claude helpfully deciding to pick up some tasks along the way.

I feel like most of my conversations with Claude lately are a battle of “I am a technical user, share technical details, but not random superfluous gibberish”

There are differences in labour and environmental laws that I do actually want a tariff to equalize.

If we have standards and laws it’s unfair to allow outsourced competition to skirt them and collect the savings. That’s a recipe for disaster.

I would be lying to say this doesn’t piss me off. I’m very angry.

My wife made a great point: this is a wake up call for me to channel my frustration into local activism. So I’ll be doing that.

If any other angry Canucks are reading this, consider doing the same. Use that anger to make our beautiful country stronger.

It’s definitely a heavy thinker, like most Qwens.

I see strings like “write, now.” In the thinning traces then it goes on to think for a lot longer, so it’s kind of weird.

I haven’t figured out hope to use this effectively yet on my strix halo.

Donald Trump has said exactly this, repeatedly, for two years straight.

He posted a picture of him taking over Canada and arresting Carney today.

That is not propaganda. And if you say it's "joking", if you threaten to shoot your friend multiple times, you'll find they don't consider it a joke.

We should still expect significant performance improvements.

I believe Unsloth’s branch (and ggml’s) don’t support MTP or the ngram embedding yet.

Also did not expect this. I did expect MotherDuck to acquire them at some point (in fact, I kinda thought they already did!)

It is under-utilized and not getting the attention it deserves.

It probably means nothing, but it’s now impossible to rationally say Canada is acting unjustifiably or targeting the US to legitimately trigger these new tariffs. It’s like a PIP for our relationship with the US.

However the US government is not rational anymore, otherwise this will get challenged and thrown out. That’s a long shot though.

As a Canadian, I’m so tired of the US’s antics. Everyone I know is. Sun Tzu teaches to never fight an enemy with their back against the wall, and I think the US will find they’ve emboldened Canadians to absorb pain as there’s no escape.

The only solution I’ve found that works is asking Mistral medium to rewrite all of Claude’s documentation and comments, then I review and rewrite the final draft for anything mistral misunderstood.

I find Claude has become very difficult to work with and incapable of writing clear documentation, even when directly prompted or provided samples.

As for code, I think each function requires 3-4 passes with Fable to actually get to a point I accept as good code. I am picky though.

The other Claudism that drives me crazy is when it writes comments and commit messages that track how you arrived at an decision instead of what it is.

There’s a good Star Trek TNG episode about this: When the Bough Breaks (S1E16)

I always wonder if minimum wage does more harm than good.

This conversation usually derails, so to be clear no, I don’t think people should work for poverty wages, a viable business should include happy non-subsidized employees.

My question is whether a global price floor psychologically depresses wages by establishing a global price anchor. Price anchoring is well known to influence the price people will accept.

For jobs like cashier that are fungible just blanket get minimum wage, would they pay more without one when the employer can’t just point to the anchor price and say take it or leave it? Would they be faster to localize wages to local costs?

I use this in a customer facing application and Gemini’s speed makes the experience feel much better.

The application isn’t so complicated that you need opus level reasoning or code writing, we need “good enough” data retrieval and processing with natural language queries and the ability to answer follow up questions.

For that Gemini works well for a decent price.