nevi-me

I cancelled my Claude personal sub and still use it at work. I tried Codex out cos 5.6 Sol, and when I gave it a security task; instead of downgrading me, it flat out refused to continue.

That left me curious that there might be a security issue that it found. I'm setting up a sandbox so I can use Kimi to see if there was an issue to be found.

I maintain a transit website as a hobby, and I'm building a Flutter app, from scratch. The old pre-COVID one carried mental baggage.

I spent a few weekends building comprehensive plans, designs, user maps, etc with Claude. So it has enough context to make decisions and keep going.

One session lasted over a day, I imagine partly because Fable + superpowers feels slow. I have an app on my phone that I have been test running since Monday on the bus.

What really helps (not sure if Opus used to do this) is that Claude will run through the emulator on its own, verifying that the design aligns with the Figma design system we created.

This is all building on top of 15 years of existing backend and rich features, so it's not a "build me a transit platform from scratch" where AI can end up making bad decisions.

I maximised usage and have reached the limit. I feel like I did 2 week's worth of hobby work over the last few days.

I got Fable to write multiple plans, spent a great part of the weekend reviewing them. Then with superpowers I left must of those plans executing with little intervention over the past few days.

I struggled to get Opus to just keep going without trying to convince me that it's late.

I found it slowed me down significantly at first, and produced more verbose code. After a few weeks of using it, I think I've gotten used to it (sometimes I explicitly bypass it, but it's good enough to know which skill to use).

Yeah on the token consumption, I'll be doing something small at work, and it'll consume a lot of tokens.

They should give us a month of access again, I feel like I didn't do enough with Fable before it got blocked.

I only realised late that I had an algorithm problem that existing models were struggling with, and Fable had made progress with. It created a 14 phase plan, which I was able to execute with Opus after the restriction.

I presumed that as a kernel developer, he would run the kernel he runs, which would require rebuilding periodically. Daily doesn't make sense, monthly is too infrequent given the rate of change in the kernel.

My speculation though. When I was building an app I was using, I used to run a recent stable build on my device instead of just the one released in the Play Store. Simplifies having to keep multiple devices.

Does Docker have uarch level support? I think similar to arch level, it could be beneficial being able to pull a v4 image.

Ubuntu started allowing defaulting to v3 packages, and I opted in. I already use the -C native to enable AVX512 when compiling binaries for local use. This matters a lot for compute/analytics workloads in my experience.

They should just license the tech from Google. Google might be missing the urinal with forcing Gemini down everything while not improving basics, but their keyword detection remains good for me.

Most of the code is already in libraries. rs-matter includes a lot of batteries, so for simple clusters you end up writing little code.

Is it their first EV? I presume the tech is outsourced or bought from competitive players that have put in the R&D. It feels like buyers will be buying the brand.

The Mercedes GT EV is faster than it, so the performance doesn't stand out.

What's missing is changing the stallion to a kiddified pony, to match the rest of the design.

This looks like a child's toy.

The keynote was the most boring for me. I paused it to go to the bathroom, and even forgot that I was watching it.

I think they've lost track of the meaning of IO and its keynotes for users. They should rather have a separate Gemini event like they had a separate Android one last week.

They're collectively losing track of their product verticals because they're too focused on shoving AI down everything. Google Home is a cluster-f, basic things keep failing, and every other announcement from the Google Home VP is about Gemini. It took them years to reduce the frequency at which devices go offline.

Even their sessions seem underwhelming. It's a mixture of "what's new in X" and "AI" this and that.

It can be the difference between between feeling like you're suffocating, not getting enough oxygen to rest enough/sleep well.

I notice a difference if I move between a ventilated room vs congested one. I suppose it depends on what's causing the concentration. If it's human breath, I'll smell freshness. If it's e.g. burning a portable gas heater (common in my part of the world), I'll feel like I'm not inhaling smoke (probably small amounts of CO).

A few years ago, I would sometimes wake up at night and open a window wide, or go open the outside door and stand there for 5-10 minutes.

It seems fine if you express what you did without focusing on the code.

It resonates well with what some people have been saying about building software for 1 person.

That CO2 concentration looks unhealthy, I wonder to what extent it's affecting your sleep quality (as opposed to waking you up).

Measure before you fix

In my case, I got a few IKEA CO2 sensors, and after leaving them in the bedrooms for a few days, we found that leaving an outside window slightly open + the bedroom door open, kept the CO2 levels below 600PPM at night.

We're 1000ft/300m away from a motorway, but fortunately the noise pollution isn't bad. So ventilating (even as it's getting cold) turned out to be a simple fix. I hadn't thought of collecting sleep data from our devices, but maybe I'll get an AI to do that, so I can correlate our sleep quality with the environment.

These tools and the analyses they have done have triggered somewhere between two and three hundred bugfixes merged in curl through-out the recent 8-10 months or so.

If you've just gone through a lengthy analysis of your code with other AI tools, surely it's reasonable not to expect to see hundreds more from a new tool?

It should be possible, unless more bugs are introduced, to eventually get to a state where there are no more bugs in your code.

Process aside, it sounds like Daniel expected to find dozens/hundreds more bugs.

Perhaps not a good example, I tried running local models a few times, to much disappointment (actually made me skeptical of LLMs in general for a while).

My last experiment in January was trying to run a Qwen model locally (RTX 4080; 128GB RAM; 9950X3D). I must have been doing it extremely wrong because the models that I tried either hallucinated severely or got stuck in a loop. The funniest one was stuck in a "but wait, ..." loop.

I fortunately had started experimenting with Claude, so I opted to pay Anthropic more money for tokens (work already covers the bill, this was for personal use).

That whole experience + a noisy GPU, put me off the idea of running/building local agents.

Reminds me of our President (SA) asking his colleagues to sign a pledge not to do corruption; when there's already a swearing in that does the same thing.

The guide is very detailed, so I've saved some for later. I've been using NATS I think since around 2018-2019 (can't recall). I've only used its pub:sub feature as it was MUCH lighter than Kafka.

It's interesting that the platform has grown so much. I paused reading at the inbox feature, so there's more to dig in to. I enjoyed reading the topic guide, and I think it was pretty intuitive when I started using it.

Outside of work projects, I maintain a public transit info site, where I either estimate or process telemetry feeds to generate trip updates, alerts, vehicle positions. NATS pub:sub works so well for me (along with tidwall:tile38 for geofencing). The site isn't that large, but a large volume of messages pass through NATS every few seconds. It's really been a great reliable small piece of technology.

To an otherwise defenceless country, it's really the same thing. Indiscriminately flattening buildings without notifying civilians to move, destroying industries, stealing their resources and reserves.

Who can recover from this, especially a small nation? You might as well declare everything to be radioactive.

So they'd react harshly even when they started it.

The norm in other countries is 6 months. That's enough time to get the mid-year numbers to be reviewed by an auditor.

I don't think malice of the decision.

The second port is likely necessary for USB hubs that rely on both ports. I had one for my M1 Air. I assume it'd still work with the 2 different speeds, but I'd be curious to try it.

I'm going to get a Neo for my wife once it's available in my country.

I'm a bit upset that there's no screenshot of Africa in the call outs at the bottom.

With the detail spec that the author describes, it reminds me that I have an identical CPU but I couldn't get my RAM to run at the advertised 5600Mhz. Hopefully there's updated BIOS so I can try did the issue again. Anyone know if I'd notice meaningful difference by flight from 3600 (what the pc reports) to 5600Mhz?

My 2017 model probably short circuited during a lightning strike, because it stopped working after a storm. My friend offered to sell me his 2019 but I thought there'd be new hardware. I should have bought it.

I love the Shield, compared to even the Chromecast at the time, we noticed a huge difference in colour on the TV. If NVIDIA ever produce a refresh, they'll have my money.

We believe that this is the largest rollout globally of any library written in Rust.

I suppose this is true because there's more phones using WhatsApp than there are say Windows 11 PCs.

Given that WhatsApp uses libsignal, is it safe to assume that they haven't been using the Rust library directly?

I've been a skeptic, but now that I'm getting into using LLMs, I'm finding being very descriptive and laying down my thoughts, preferences, assumptions, etc, to help greatly.

I suppose a year ago we were talking about prompt engineers, so it's partly about being good at describing problems.

It's not just tax obligations, no? Employers in many countries have an obligation to ensure that your salary reflects on the X day of the month (or whatever frequency you're paid). Banks in my country have a payroll payment system for this reason, where funds will clear on the day they're made despite the destination bank (in the same country).

If my employer has to use SWIFT to pay me, on whom does this obligation to ensure I'm paid on time fall? I've had a salary payment from a foreign employer fail to be delivered for 2 weeks a few times. We'd have to go back and forth with my bank, their bank, their payroll vendor. That's an exception because they hired me as a foreign employee. Despite paying their local employees on time, I always received my salary at least 4 days 'late', as long as their payroll system reflected that I was paid on the X day, it wasn't their problem.

I somehow thought the launch of the car was next week, I see it was this week.

I thought I'd wait for it but my ICE started giving me mechanical issues that weren't being resolved. With this and the BMW iX3 being around a year away from my local market, I ended up getting a PHEV X5.

The Volvo looks shorter than the XC60, looks more like a station wagon than an SUV. I'm only on my 3rd car so I don't have experience/knowledge to understand what people mean when they say Volvo is no longer what it was.

On the 'coffee shop charging', I don't really prioritise quick charges because when I make trips long enough to require pit stops, those stops are normally a chance for the kids to play and eat.

Performance-wise, I don't mind 0-100 in the 4-5 seconds range. I test drove the EX30 and accelerate sharply from a stop. My wife complained about the whiplash, so I imagine it would be dangerous to restless toddlers, as they already complain about the X5.

Lastly, V2L is welcome, the range is good (for me) for the battery sizes, but it looks like the iX3 would be a better car for the larger battery. Tangentially, BMW claimed that the iX3 would set a new benchmark in EV efficiency, yet Volvo is claiming similar ranges with a 10% smaller battery.

Time will tell on their battery, especially if the bike they're putting it on delivers. I think the overall point could be that there's active R&D in trying to find geopolitically sustainable materials, and lowering the weight of materials used.

I sympathise with burying cables that you think are for life, only to need to replace them later.

We ran fibre cables in the ceiling when constructing our house. I requested the electrician to shield the cables with some tubing, but he probably thought I was being extreme. We have 9 cables, 2 of them don't work, likely from being bent by mistake or something.

The wiring is intermixed with electrical and ethernet (for cameras) cables, making the process a bit tricky. At least for us we might only have to cut the ceiling boards in a few places to help guide the replacement cables.