hadlock

As long as it was the kind I could buy spare parts for, sure. And it wasn't connected to the internet. Not having cleaning staff in the house periodically would be a huge quality of life improvement.

I suspect in 5 years everyone will have the equivalent of a 512gb mac mini running a 500b class open weight model for 98% of their tasks, shelling out to openai/anthropic for the other 2%. People who need higher end models will own the equivalent of 4 x 512gb mac mini running a 2.8T class open weight model. I don't know where OpenAI and Anthropic are going to get their revenue from to keep developing SOTA frontier models at that point.

I've had 2b models give a plausible Paris vacation itinerary. A tools-capable 12b and especially 30b model from 2026 is certainly capable of producing passable results. I was demonstrating the qwen 3.6 27b model I stood up last week to my wife and it gave her a passable Moroccan Chicken recipe. With tool calling (search) they're quite good.

The big problem with US AI is that they deprecate their models after like a year. If I have a routine business process that works with GPT 9.9 and then next year they release GPT10 and 9.9 isn't available anymore, I really do not want to have to drop everything and verify that 10 behaves close enough to 9.9 for my specific task. With an open model I can just host it on whatever hardware or cloud instance forever. Most software you want to keep up to date to avoid security issues but with an LLM you can update the harness and keep the weights forever.

When I got married our food waste probably tripled. My food waste as a single guy was significantly less than married. When I had kids it probably tripled again. Much like a horse to water, you cannot explain to a 3 year old that if they don't eat all the banannas before tomorrow, they will go bad. We probably throw out 1-2 lbs of fruit a week, which is ~100 lbs a year right there.

The kid might have ate 2 lbs of blueberries a week for six months, but surprise, they hate them now!

I can't count the number of half-full glasses of milk I've found in the last month, probably averaging 1 per day. Kids are extremely inefficient eaters unless it is covered in sugar or chocolate.

When my wife finally upgraded her iphone to one that uses the usb-c, finally the last non-usb-c device in the house was gone. It has drastically simplified our lives. Each car only has a single usb-c cable now and that's enough for long road trips even. We got a set of Anker cables ~5 years ago and they still haven't worn out yet

You are not going to get the guy at 7-11 or the cashier at Target who just bought a PS5 for her son to boycott watching movies on it. Boycotts only work if it is demonstrably going to make their life worse if they don't. Losing access to a movie that interested you 15 years ago when you were still in high school is not one of those things.

I'm currently supporting (a couple hours a week) an exec who has mostly automated his workflow via vibe coding (the rest largely failed - not everyone can make the leap, and those that do are largely marginal) his workflow is largely a VS code window with the git commit/sync button on the left, and 2-20 codex/claude tabs in the main window on the right. I think he could actually make use of this sort of thing. Extremely small sample group but I'd estimate 1/300 people in a tech centric location like downtown SF. Globally it would be a tiny tiny fraction of that number lol. The venn diagram overlap of people who this would be useful for, vs people who have already vibe-coded their own macro keyboard with a streamdeck (myself included) is probably a pretty strong overlap. It's an amusing marketing gimmick, and historical artifact, if nothing else.

Great question - nothing. I was a little surprised to see it doesn't even have a microphone or speaker - that means regardless of whatever compute is available (esp32 is pretty common in the $30 price class for smart assistants) this can never be a standalone device.

The J stands for Jacobian, ref Jacobian Matrix, Jacobian "Lens"

If the app store allowed you to install bookmarks, I doubt most people would notice.

The other big problem is, group notifications went from 100% success rate, to nearly 0%. I bought a budget model sainsmart/genmitsu cnc thing to goof around with, as a hobby/assist other hobbies, and 98% of their user community is on Facebook, but Facebook doesn't show me their posts most of the time, because sainsmart isn't paying to promote their group. This group would be a wealth of information every time I opened the fb app, but unless I explicitly remember to go visit their "community" by searching for it, I'll never see those posts. There's an enormous Cyclekart (vintage-ish go karts with proportionally correct wheels) community on fb as well but unless that's the ONLY thing you use their app for, they won't show you that stuff.

TL;DR FB is worthless for social networks, because if they showed you the groups you were there to check on, nobody would pay to promote their own things.

Once an avid user, I go 1-2 weeks at a time with the app uninstalled now unless there's something specific I need to check on.

It means a lot more if you held one of these in your hands new before you were old enough to drive

(or, in my case, read about it in nintendo power, but only saw it behind the glass case at Target)

so they can be "skilled artisans"

Having had to work with these guys, and then maintain their software when they inevitably get bored and/or leave for more money elsewhere, no. Usually when these guys leave, their stacks/projects are the first to get rolled into the monolith and/or rewritten in the company's lingua franca (python)

Apple rumor mill is suggesting that we may see M5 Mac Studio announced in September at that event. Apple just increased their pricing across the board, so I'm not holding my breath that they will be reasonably priced. They also cancelled development of their M6 in favor of the future M7 which I suspect will be a major, AI-focused upgrade compared to the M4->M5 upgrade

it uses JIS screws rather than Philips (in addition to triwing)

I don't think this is an issue for anyone who has had to disassemble a japanese device before, and the bits are widely available online. Countless youtube videos have discussed that JIS vs Philips in the consumer space are largely compatible outside of american aircraft construction.

They added a "/goal" command which I guess spawns a supervisor agent process that checks to see if your goal statement has been achieved (e.g. "/goal complete tasks 1-250 of plan.md") I've been pretty happy with it but I rarely use that workflow. Most of the time I give it a 3-6 step prompt and come back in 20 min and the first two were done and I get a summary "up next is to complete the next steps" which.... Opus 4.6 didn't have this problem. 4.8 feels like a cost cutting measure, or maybe it's just tuned poorly for my specific workflow (multi-repo system integration)

whoops sorry, I just made it public. long holiday weekend

In my subjective experience, Dextromethorphan (DXM, as the robo-trippers call it) does almost nothing for me, in the 1-5% range

The only cold and cough medicine that really truly works is the over-the-counter stuff, pseudoephedrine, works amazing for me. I usually pick up a box of the stuff when school starts in the fall and I go through half a box of it by the following summer.

I ran W2K through most of high school and until like 2009 when Valve finally dropped support for it. It was a great OS fast, rarely crashed, most games would actually run on it. Valve dropping W2K support meant TF2 no longer ran without jumping through a bunch of hoops

One of the key design principles of podman was rootless operation; they were so disgusted by Docker being a daemon they decided to do a full open source implementation. I've never had an issue with it running without root.

Here you go. The green bar is the claude code 5 min cache expiry countdown timer (anthropic charges a fraction of cached context, I forget something like 10 or 30% of the normal price, but the cache is only 5 min for most users, 1 hr for enterprise/premium)

https://github.com/Hadlock/cscope/

It's not amazing code but it's tooling that works for me.

I have a pretty specific workflow, I deal with live systems a lot (primarily in staging but occasionally prod) and have a really neat skill that has hooks into all the various things and then writes (&& checks in git) an investigation doc (.md) using a template into a git knowledgebase by month/quarter. That knowledgebase is really useful for frontloading the context of my new sessions as a lot of my work is iterative on previous work. It is really good at 80% of tasks, but simple stuff like using sops or modifying subcharts in helm it really struggles with. I do a lot of integration work between monorepos and Opus 4.7/4.8 really struggle with that kind of work. Opus is so bad at sops work I just outsource that to Codex/GPT5.5 now because the failure rate is so high.

The other high failure rate things is opus correctly predicting where endpoints will end up (stacked endpoints), when you have an ingress with a main monorepo but it's endpoints are the outputs (not just json regurgitated/passed forward).

I figured by having the clanker write it's own documentation about what it did, and then the next time saying "go look at invest doc 2026-06-29-TICKT_1234.md, do that again, but here's what we're doing differently and why..." and it would immediately grep the situation, but it doesn't. As alluded to earlier I've been exploring other models since the corporate blessed one is struggling to keep up.

Curious to see what you release.

I ended up building a vs-code/IDE style workflow for claude that has a "file browser" of CC sessions in the left column, sorted by repo, and then terminal in the right column, and then I've been tacking features like this onto it

Opus is so slow these days what I really want is a bell sound to ring when it's done. I kick off some task and then it takes 3-12 minutes to complete. It's wrong, so I tell it to revert and try again with slightly different instructions.

I guess if you enjoy engaging with an echo chamber

This is the correct answer

Welcome to using v1.0.0 of any product

There was a flight to quality (their first real test, in my opinion) back in February 2026 and... nobody bought Crypto. Crypto never recovered after that. It's a good time to be a lobbyist, but I don't think you can effectively legislate against the value of gold for any amount of money.

making software better instead of banning it

We're still in the middle of the cambrian explosion.

If Anthropic was capable of developing Opus 4.49-4.5 2H 2025.... then any company with a research team capable of reading all the papers and press releases will be capable of producing Opus 4.8 by the end of 2027, either raw model competency, or in a harness like claude code (or better with both). I guess what I am trying to say is that Opus 4.5 does not represent the edge of agentic capability, merely somewhere in the thick meaty layer of "functional and achievable".

We can draw the line at Sonnet 4.6 in the US but much like encryption export restrictions in the 1980s, the line drawn will be laughably low within a few years and simply unthinkable in a decade.