8,795 karma · joined January 1, 2016
Looking forward to seeing some benchmarks!
For sure. Still much cheaper than vacuuming every day.
> getting stuck
I appreciate that this is a function of the specific environment and robot you buy, but as a personal anecdote, this doesn't happen to me in practice. There was a setup cost, but once that was done, it doesn't get stuck.
> Emptying the vacuum bag
Not that frequent at all, considering that daily vacuuming keeps things remarkably clean and thus little is added to the main bag in the docking station each day. This is also far less work and far more preferable than doing the vacuuming myself.
The analogy is like pointing out that you can have a robot vacuum clean your floors every day. Before the robot, you'd do it maybe once a week, just no time - "too expensive" to even think about doing it daily. The robot makes it happen all the time for a tiny, tiny fraction of the cost.
The question of whether it's better to frame a reaction to issues in a positive, do-better-next-time or a negative, don't-do-that-again way (fwiw, I agree that the positive way is better) is orthogonal to the argument I was trying to make about who is the recipient of that reaction. Scapegoating the agent is farcical. Clearly, "make no mistakes" style prompts are a poor approach to agentic development.
> the manager's job is to enable career progression of their reports... we aren't enabling the career progression of [fancy calculators]
While I do agree that good human managers concern themselves with career development of their human direct reports, I disagree that it's a fundamental part of the job description. The foundation is for a manager to get and keep their reports being productive. Human workers are often concerned with career progression, so good managers concern themselves with such career progression as well. Agents are not, but that doesn't mean that lauding good agentic work is meaningless. Lauding the work is also valuable for: (a) reinforcing a culture of positivity and celebration, which rubs off on human workers as well, (b) part of a culture of continuous evaluation of agents and models - asking questions like, are there cheaper or faster models that would be just as effective?
A bad manager takes personal credit when things go well and finds a scapegoat when things go bad. A good manager gives credit when things go well and takes responsibility when things go bad.
It's wrong to scapegoat the agent. You, the human who executed the agent, made the decision to execute it. You gave it credentials, and set up insufficient guardrails. Don't scapegoat the agent for your poor decisions.
Likewise, it's good to have humility when you get great results out of agents. Yes, the agent made the good results happen. It's OK to tell people that your choice of agent is amazing and did an incredible job. In any half-decent work culture, that should rub off on you as well.
They can be both great and over-valued at the same time.
I've worked for a boss like this once - he was the worst boss I ever worked for. There was absolutely no room for nuance in the conversations, no attempt to build rapport or a sense of shared understanding. It was his way or the highway.
Human beings are emotional, not rational actors. The "idle chit-chat" gets the foundations of emotional solidarity in place that make anything else possible. When you try to avoid it, you get an organization slinging daggers at each other because someone had the nerve - the audacity - to tell people what to do over an email without any room for them to respond and feel heard, never mind that, rationally, that's exactly what they were hired to do.
There is, of course, value in getting to the point - but only after you consistently build and assure and rebuild and reassure confidence that everyone is in the same boat together.
1. Homebrew really, really wants you to host your taps on GitHub. The docs are very GitHub-focused: https://docs.brew.sh/How-to-Create-and-Maintain-a-Tap as are downstream projects like dist: https://axodotdev.github.io/cargo-dist/book/installers/homeb...
2. Terraform and OpenTofu public registries support only GitHub, see e.g. https://developer.hashicorp.com/terraform/registry/modules/p... and https://opentofu.org/docs/language/modules/develop/publish/#...
3. Only GitHub allows you to use free minutes on managed Windows and macOS runners, which are essential for building and signing native Windows and macOS/iOS software. GitLab's are in beta and require a paid plan, Buildkite requires you to pay at least $30/user without any bundled macOS minutes and no hosted Windows runners available. And for small shops, keeping the macOS and Windows runners clean is such a stupid time sink, with the small number of minutes involved it's much more preferable to just start with free managed minutes and plan to pay for managed runner minutes later.
OpenAI thinks their API is really worth more than 2x the price?
Two things are true at the same time: that as an employee, you will always get a distilled picture of the wider reality from your manager (sometimes to help you focus, sometimes because they're forbidden to share it with you, sometimes because parts of it would damage morale and thus productivity), and that managerial competence is in the communication skills to achieve that focus and morale. Two different managers with different skill levels, each dealing with the same "full facts" reality, will deliver different results.
Related tangent: it's sad to see the decline of high school debate teams, there's probably no better early-life preparation for the management skillset. Competent managers persuade. They know who they're trying to persuade, what that person values, and frame the wider organizational reality in such a way that it will make sense and align their direct report.
The worst managers think their job is to tell you what to do and that your job is to do what they say.
I do think agricultural subsidies are distorting the market and our food system, but eliminating them isn't going to magically fix everyone's diets overnight.
Wrong. As food gets costlier, people shift eating habits from healthier but more expensive food to cheaper ultra-processed food. There are similarities elsewhere, like the iron law of prohibition: making opium more difficult to sell shifted the market to heroin, and eventually to fentanyl.
Is actually a really good example of why TS was a bad choice for Cloudflare. It made sense for their initial version, because the only programs that you could run on Cloudflare Workers were themselves JavaScript scripts, so it was a reasonable expectation that developers were already using npm/yarn/pnpm to install tools and manage dependencies.
But now, Cloudflare Workers supports containerized workloads and Python workloads. So even if your project doesn't use TS, you are still expected to set up some kind of JS toolchain for your project, just for wrangler.
Choice of tech stack is not merely a question of performance, but also portability and packaging. Self-contained, compiled static binaries are much, much more portable than shipping full script runtimes (JS/Python), dealing with the inevitable spread of runtime versions installed on everyone's machines, and dealing with the version spread of the dependencies installed separately.
> All the CLI does is take command line arguments, put together requests based on them, send them to an API, get the responses, do mild transformation on the responses,and output it to stdout/stderr.
This is an argument in favor of a language like Go, arguing that the garbage collector in the runtime is inconsequential, over a language like Rust. It is not an argument for shipping a script runtime.
Not my experience with Claude Code.
> writing a deterministic, traditional CLI tool to poll GitLab CI pipeline+job state changes on a branch and exit with an appropriate status code
This is what Claude Code does, more or less, on the fly. With a short prompt like "I pushed, monitor CI and debug if needed", it writes a monitor script which is responsible for polling CI status (the script is short, so it's not token-heavy), and if CI fails, only then does the agent proceed to pulling out CI logs, grepping them for signs of errors, etc. as continuation to debugging.
I mean, I'm sure it's more token-efficient to have a CLI tool ready-to-go instead of Claude Code dynamically writing its own script each time, but as I'm on a Max sub where it doesn't seem to affect how close I am to the limits, and I only ever hit the limits if I'm running Fable for everything... /shrug
I live in a place with year-round fair weather, so I use a petrol motorcycle as my primary vehicle, and live in a rented apartment with no dedicated parking space or possibility of using an at-home charger.
Even with incredibly high petrol taxes causing petrol to cost more than 2.2 EUR/L (like 9.5 USD/gallon), the public fast EV charger rates available at refueling stations are so expensive, that when you take great motorcycle mileage into account, it still doesn't pay to buy an EV, even when only considering charging/refueling costs. Of course, if I could recharge from slow household electric power, it would be a different story. But the politicians are already worried enough about the loss of petrol taxes that the odds, IMO, are quite low of EV charging staying cheaper than petrol refueling (particularly when the vast majority of the petrol price is taxes to begin with).
The Hobbit was published in 1937 and The Lord of the Rings was published 1954-55.
If the kids can't develop an emotional connection to them, that says much more about their teachers than it does the children or modern times.
On the contrary - it's trivial for controlling shareholders to direct the company to issue more stock to them and thus dilute the shares that were taxed. What do you propose, that companies can no longer issue stock after some of the stock has been taxed?
And what about non-divisible assets like real estate? There's nothing that prevents the government from forcing real estate to be held by LLCs instead of individuals, then shares in the LLC could slowly be taxed by the government. So what happens in 20 years when the government owns 50.1% of all the LLC shares that comprise the ownership of the $100 million Hollywood mansion? You're going to let a bureaucrat kick out the A-lister who lives there and put it up for auction? Attempt to sell it to someone who knows full well the same would happen to him?
It's very, very hard to design a wealth tax that doesn't end up being an assault on private ownership in all forms.
C4 gives the text-based representation needed for LLMs to generate and maintain large architecture diagrams - why not build off that heritage?
Who says that the price you paid per share was the actual market value of the shares? For example, let's say that you inherited 10 million options to purchase Microsoft stock at $1/share, and in so exercising the options (by writing a check for, say, $10 million to Microsoft), you then end up with 10 million shares, which on paper, with the current stock price close to $500/share, would be worth close to $5 billion. But could you actually get that much money from selling 10 million shares? Definitely not overnight - so many shares getting dumped on the market at once would materially affect the stock price. The $5 billion number is a hypothetical that depends on other people backing up the hypothetical numbers with their own money (i.e. buying at the hypothetical price) - it is not the same as "I have $5 billion in a bank account and could use that to go buy a yacht and buy political ads etc. with it"
This makes it exceedingly easy to ensure that old people are not affected by property taxes: if you are elderly resident, you simply get a discount. If you are an elderly landlord who owns multiple properties, your younger tenants don't get a property tax discount, and you pay income taxes (not discounted for age) on the rental income. If you are elderly and want to downsize, your property tax discount follows you into the smaller apartment.
The system has its challenges, but it is far superior to Prop 13 and achieves the same goal.
Docs do not.
So what results is either allowing developers to push (but not force push) directly to main (you can always push revert commits if needed), or a PR process that exists to enforce linters and build-ability, but if those pass, allow the developer to merge independently, without human review.
Companies feel like they have no choice, for two reasons:
1. Jurisdictions where workers must give minimum notice periods before leaving their prior employer - the worker accepts a competing offer, gives notice, waits out the notice period, then starts. And many workers often ask for additional unpaid time between jobs before starting (since no vacation days have been accrued yet at the new job).
2. Hiring happens by committee. In most companies, job candidates must be interviewed by, at the absolute minimum, an HR screener, the hiring manager, and the hiring skip-level; and usually also by people who will be peer teammates, plus an HR "culture" interview. If you need to pass a committee of 3-6 people, interviews must be scheduled with each member of the committee; at one committee member per week, that is 3-6 weeks.
6 weeks (1.5 months) plus background checks plus getting pay packages approved plus 30 day notice periods plus two weeks unpaid vacation is, yes, a multiple month endeavor at minimum to fill a role. The only way to fill a role within a week is to grant full hiring autonomy to the hiring manager and only hire people who are currently unemployed (and therefore have no notice period or desire to take more unpaid time before starting), which in most companies beyond early startup size, is a complete non-starter.
I agree with you, but it's worth spelling out exactly why. Philosophically, prison has one of three purposes: protecting society, rehabilitation, or punishment.
If the sentence for use is anything less than spending the rest of your life in jail, then society isn't really protected from addicts; the recidivist rate is too high. Get out, steal again for the next high, rinse and repeat through the revolving door.
Prison is not an environment capable of rehabilitating addicts. Leaving aside drug smuggling into prisons, getting people past the withdrawal period is the "easy" part of rehab, the real work is in therapy, developing better relationship models, better boundaries, better strategies for the triggers that drove people to look for a high. It's very, very difficult to facilitate that in prison.
Finally, it's debatable whether prison is actually punishment for addicts. If they keep access to drugs through the people who smuggle them into prisons, and the prison pays for meals and a roof that were often lacking for the addict on the street, then, is it really a punishment?
Finishing your goals does not "make you right". Life is not a binary where you are either wrong, or right. Thinking that you ever become "right" or that you ever achieve 100% confidence that you have arrived at a "solution" is arrogance of the highest order.
Life is about getting better. Every person, every product, starts at a beginning. A beginning is not "wrong", it is early. We try, individually and collectively, to improve ourselves and that which we are responsible for. The decision to ship is not whether we have made something "right" but about whether we have made it better. We ship, knowing we will ship another improvement after this one.
> 2. Gather missing information... 6. Achieve the goal
Sometimes in life, you need to ship first in order to gather missing information later. You cannot do that if everything has to be "right" before you ship it.
> I love being wrong
This is not the essay of someone who loves being wrong. This is the essay of someone who loves becoming right. But the deeper truth is that you will always be "wrong" - unfinished, lacking knowledge, under-developed, incomplete. Do not love being wrong - love improvement.
60/hour sucks. 5000/hour (a little more than one per second) is totally fine.
I'm chalking this up alongside Docker's decision to restrict unauthenticated pulls. Unauthenticated anything went the way of the dodo some time ago. If you want unauthenticated access, go run your own mirror.