HN Simulatornew | past | comments | lists | submit | colinb's commentslogin

I was curious about the numbers. Apologies for the year mismatch. I can only report what I found. This year there are about 466000 teachers working in primary and secondary education in the UK. In 2024 there were 1684 complaints to the TRA about teachers. (No idea how many were to because of physical contact) of 1055 were dismissed. So around 600 upheld complaints. Or 0.3% of teachers. That doesn’t seem too bad to me.

The greater scandal is that the average time to clear each case was around 103 weeks.


>1000 referrals annually is a lot and can be damaging to ones wellbeing and lead to a dismissal even if it doesn't lead to a suspension. Most teachers are not men (only 14% in primary, 35% in secondary) and are almost certainly not doing what the above suggested is allowed.

It’s less than one per hundred teachers and that’s with the assumption that there’s never two complaints about a single teacher in a year. I guess we have different ideas of a lot, which is fair enough.

I have no direct experience of this being an engineer rather than a teacher. I have been subject to an HR complaint once in my career. At Cisco. I’m not sorry. I can easily believe that it’s a stressful and unpleasant experience, especially for those people, the great majority, who are exonerated.


I only consider it a lot because that's usually the final point of escalation after it goes through safeguarding or a formal investigation conducted by the school before it reaches the government. But it doesn't have to reach that point to hurt your career. I just know enough from teachers that I wouldn't try and practice what OP suggested.

To be clear, I'm not suggesting it, I'm sharing what the law says. FWIW, I am a qualified teacher and a man, if those things are important to the discussion.

What the law allows may still fall well below the professional standards expected of a teacher. In particular, a teacher who uses force to maintain good order and discipline is likely failing to create a safe environment, failing to manage behaviour effectively, failing to treat pupils with dignity, failing to safeguard children's wellbeing, amongst others.

I would fully expect a teacher who can't control their classroom without force to be investigated and to potentially lose their job pretty quickly. The point is that it's not illegal, provided the force is reasonable (another point of ambiguity).

Also, it's not 600 upheld complaints, it's 600 investigations resulting from referrals. After investigation, 244 hearings were held and 202 of those found serious misconduct, and 157 teachers were banned from teaching, most commonly for sexual misconduct or indecent images, not use of force.

466 000 is FTE for state-school teachers in England alone. Actual HC is 512 000, with another 75 000 FTE working in independent schools (couldn't find HC details), so 600 000 is a better estimate for the number of teachers under the jursidiction of the TRA - 202/600 000 is 0.03%, or one in every 3000 teachers (referrals is still around one in 350).

Obviously stuff handled before it reaches the TRA isn't in those stats, but mostly those cases are not career-ending.


I know. I’m willing to bet the first thing they’d do after seizing the guns would be to stop unfriendly media from first hand reporting of events in the White House.

Then, when they know they’re proof against an armed citizenry, they’ll start running government policy as a way to line their own pockets. Why, it wouldn’t surprise me if a sufficiently emboldened crook, safe in the knowledge, that you personally cannot stop them by armed force, started to flout the laws left and right.

Put it another way. I never fail to be amazed by the fantasies that govern significant chunks of American thought. People who’ve lived there, and also in other places are able to figure out pretty quickly where they’re actually safer, and more respected by their government, and able to live a decent wholesome life.

I’ll note in passing that there are other countries with similar levels of gun ownership but less suicide and violence. Perhaps the first is because they have slightly more left leaning social policies. I’m all but convinced that the second is because they aren’t driven to a frenzy of terror by their local media. Maybe that cat has escaped its bag and isn’t ever going back. A shame. To live one’s life in fear of angry frightened people with guns sounds awful to me.

Last thing. I’ve had guns pointed at me twice. Once by an infant British soldier in 1990s Belfast. And once by a robbing bastard in a very wealthy part of Los Angeles. Neither was fun, but at least the soldier was predictable in the then prevailing circumstances


I, too, have been threatened by a gun. Once by a cop, the other was a mugger.


Your italicised text is /exactly/ wrong. Hen eggs are fertilised inside the hen’s oviduct. I mean, eggs are somewhat porous after laying, but they’re not that porous.

(I don’t know about you. I’m just a programmer holding strong opinions about avian physiology. Maybe there’s an actual biologist on the plane who can correct me)


Oh! Thanks for correcting me. Now I learned that the egg is fertilized (if at all) before the shell is formed. The hen would lay the egg anyway fertilization or no fertilization.


Simplified egg laying are the chicken analogue to human female menstruation. It is technically quite different (chickens don’t have a uterus, for one), but it is still analogous.


Not accessible in the EU cuz GDPR. “We thought about having a business that wasn’t exploitative and abusive of our users, and decided against it.”


I think I strongly disagree with your first sentence. Strongly. I've worked with enough people with fine coding skills and terrible English to know that the two are not an uncommon combination. Last year, I worked with someone whose first language was something from China - no idea which flavour - and they would often pass their written comms through an LLM. I could tell when they had done so by the onset of sudden grammatical correctness. But I don't think they were doing this with the intention of deceiving, merely of communicating clearly and correctly.

Now, I often find Claude's idea of idiomatic prose to be a bit load-bearingly seam-hitting as it lands not this point, but THAT one, but it is probably better than something hacked out by a person who confuses tenses, cases, pronouns and when they can enverb a noun. So I tend towards giving people the benefit of the doubt on this.

Lastly, if you think LLM-written software is full of security vulnerabilities and bugs, I have terrible news for you about the state of human-written code. The fact that what we do is often better than nothing at all, is no big recommendation.


> I've worked with enough people with fine coding skills and terrible English

What about people with fine English and terrible coding skills?

What about people with both terrible English and terrible coding skills?

> Now, I often find Claude's idea of idiomatic prose to be a bit load-bearingly seam-hitting as it lands not this point, but THAT one, but it is probably better than something hacked out by a person who confuses tenses, cases, pronouns and when they can enverb a noun. So I tend towards giving people the benefit of the doubt on this.

Personally, I find AI writing as insufferable as anyone else (though Anthropic's is particularly bad, maybe just due to my familiarity with it), but I wouldn't judge individuals for using it to make communication more readable, rephrase what they mean etc. If anything, any difficulty in reading my prose would support that.

> Lastly, if you think LLM-written software is full of security vulnerabilities and bugs, I have terrible news for you about the state of human-written code. The fact that what we do is often better than nothing at all, is no big recommendation.

I guess a lot depends on how you use the LLMs (I bet horrible coding skills coincide with horribly lazy and problematic usage of LLMs for development too), but I wonder how humans actually stack up to the slop-machine when it comes to how good or bad the code they produce is on average, since I'm sure that SOTA model code by now tends towards the upper end of that.

In other words, if a person wants to limit themselves to avoiding LLM code over fears of it being bad, they would also have to avoid any any all code written by people that essentially produced the problematic training dataset, as well as any programmers with the same capabilities (or lack thereof). You'd basically have to avoid using a lot/most of the software out there if that's your quality standard.


Can you explain more please? I recognise some Claude-isms when reading text, and I recall image generators that added more fingers than are usual, but I don’t know what the 2-d graphics on the app home page signify in this respect.


[flagged]

> They say idiotic things like, "is the whole game hallucinated?"

You say this as if you can't even faithfully reproduce text that is at the top of your context window, which would explain the rest, such as hallucinating anyone saying anything about play testing.

> angry [..] bitter [..] mediocre

I detect no trace of that in the comment that sent you, and all the more in yours. Probably a coincidence.


Not to do the <- this, thing, but <- this.

It is astonishing how many submissions on HN now have a top comment doing the AI sniff test bullshit, usually based upon literally nothing. Aside from being angry gatekeeping assholes, these people really want to show how super advanced their AI-dar is, but it's just a gigantic waste of time for everyone else and just false positive matches everything: Every comment, image, website, game or app, book, and so on, it's all AI to these people.

Many haven't yet caught on that these people think everything is AI now -- that they desperately want everyone to know that they haven't been "had" -- so people still sincerely ask "why?" like some secret tell will be revealed, when really it's just worthless nonsense.


There are two extremes though. On one side people who feel immediately offended if they believe they are served LLM-made something. On other people deeply hurt when anybody notices LLM work, and pretending it's impossible to say.

I suppose, the healthy position is not to deny reality:

1) LLM leaves its signs in texts, code, and imagery.

2) It's not necessary a reason to be outraged. Like, for a game it's not a major thing to judge really.


Ah, a made up dichotomy. Did the AI generate it for you? Hilarious stuff. Then implying that I'm "denying reality" for pointing out that people are declaring everything AI.

There are things that are obviously "AI", in a detrimental sense. But there are many, many more things where asshats doing a purity test are running around declaring everything AI.

Clean website that follows norms and expectations of affordances? AI!

Text that you spent time to try to refine? AI!

Basically anything that doesn't look dogshit, amateur hour garbage? AI!

That is where HN is at, and these "AI purity test" comments appear constantly, in basically every show HN. You aren't seeing the "signs", you're backwards luddites that think you can shame people actually creating things, and it has reached a farcical point where it's just noisy nonsense.

Every "this looks like AI" comment should be flagged to oblivion if HN wasn't basically a retirement home for industry washouts. But it is, so they always rise to the top. Every discussion ends up tied up in worthless noise where people who've created nothing try to dismiss everything.


A range is not a dichotomy (just a nerdy note). Apparently, you're so mad at me. I'm just not sure why. You can contribute to HN being nicer place by just answering what is written, really.


Please re-read the guidelines, or have your agent summarize it for you.


I think normally it's just that someone has the thought and shares it for discussion. Not every comment is performative.


Hrmmm, is your comment AI? Are you a bot? Just being honest here, but your comment and your history on AI has a real "AI smell" to it.


...but yes, some comments are performative.


I too thought of Polanski (Roman, not Zac) while reading these threads. I’m not sure the comparison or perhaps, the parallel holds though. PO’b may have been a bad husband. Or he may have been married to a bad wife. Or maybe they were both a bit rough. I don’t know an any real evidence either way, and life has taught me that it’s hard to see inside the truth of other people’s relationships. I have trouble enough with my own, and I’m present in real-time!

Polanski R is not the same thing. He drugged and raped a child and was defended for decades by people who ought to know better because of his directorial genius. I don’t dispute that genius. I just can’t watch his movies again without feeling a little sick.

All of this to say, judging Polanski for being a monster is not the same as judging PO’b for being a jerk. I reread the Aubrey-Maturin novels every four or five years and they enrich my life on every occasion. And I don’t feel even slightly bad about it.

(Disappointingly, I did lose an argument over a big family dinner several years ago when it turned it that tracklement is a neologism. Imagine my embarrassment!)


I hoped it obvious that I am not comparing being a liar and a bad family man to being a serial rapist (unsurprisingly there is quite few allegations of this exact same behaviour against Polanski). I simply used Polanski and MJ as obvious examples of A) why not separating the art and artist absolutely is not just “immature virtue signaling” B) how people loving the art are able to compartmentalise the evidence and suffering of the victims to keep enjoying the art, meanwhile making any justice very difficult.

There’s artist with whom we can “forget everything else”, then there’s not. Where one draws the line, well, I suppose that up to one’s own discretion, but I would recommend offering a serious thought on where oneself draws the line.


ah, my apologies. I read more than you wrote. I think you and I are probably more in agreement than not. I read a book last year, "Monsters: What Do We Do with Great Art by Bad People" by Claire Dederer, on this topic. Good in parts. Left me hanging. She talks a lot about several probably bad people, Polanski, Woody Allen, Picasso etc. But she never seems to hit a conclusion. Maybe that's the answer; there is no easy, pat answer. You just have to push through each case, as you find it, according to your judgement at the time.

While writing this, I dug out a Reddit thread about the book, where one of the participants (YakSlothLemon FTW!) hits us with Wagner. Well, yes, game over. I'm not a fan of that sort of (to my ears) overcooked opera, but if you like it, play it. The dirtbag isn't getting any royalties because he's good and dead.


I have to say I’m really enjoying the Eternal September of Software comments here. They make me wonder about the vintage of the authors. Was Windows 7 the first stable OS they used? Some crustifarian Debian? I don’t actually recall if the Multicians I knew thought New Jersey software was rubbish. I know I still miss the solidity and useful abstractions of VAXcluster. Kubernetes would be funny by comparison if it weren’t so tragic.

It was always better in the past. The capitalists always ruined it in the past. Only I am the true bearer of the artisanal software flame. /s


Not in any way being negative about what you’ve done, but… Mac only. Is this just because of inbuilt knowledge about file system shape or some more important Apple exclusive tool/environmental quirk?


We’re working on Linux/Windows support right now. Mac was the easiest to setup for us since that’s the machine we code on. What OS do you use?


Not OP, but I use NixOS. I will give it a try later tonight. Do you need public contribution for this?


Would love to hear your thoughts and appreciate public contribution.


Spent some time playing on it just now on Linux. I couldn't easily get the automatic ingestion or gardening working, but everything else is working okay. For my workflows I think I actually prefer wiki updates being manual and deliberate.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: