固定されたツイート
A clock, wot I did design&make, mostly wooden. Grasshopper escapement, 1m pendulum. A bit dusty, cos it's a bit of a bugger to dust.
English
Piers C. Structures (@pcstru.bsky.social)
26K posts
Piers C. Structures (@pcstru.bsky.social)
@pcstru
Philosophy made me ignorant of everything except my own ignorance. My advice would be avoid it like the plague if you want to really know anything ever again.
参加日 Ekim 2010
2.3K フォロー中1.9K フォロワー
@nicbarkeragain I used to have dreams in code. C++ was ordinary nightmare territory, an Assembly (PIC Microschip, usually 16F84's) dream took over where that left off. Let me always dream in Pascal.
English
The entire idea of AI code generation is based on the assumption that a "prompt" can express your technical intention. But as soon as you start to get good at programming you stop thinking about it in english and instead switch to some abstract mental representation.
English
@BarbaraSutton15 @GMB @GBNEWS Funny how like a synchronised dance troupe, our journalists pivot from "How can we possibly afford that" for any public spending that benefits people to, "more taxes for bombs please", citing military bods who will only ever want more.
English
@codevsdev I'd start with code that works. No one is going to be maintaining anything if it doesn't do the business.
English
Piers C. Structures (@pcstru.bsky.social) がリツイート
The whole world is celebrating the World Cup while Israeli occupation forces are attacking Gaza..
This is for this night’s Israeli attacks in Gaza
English
Piers C. Structures (@pcstru.bsky.social) がリツイート
🚨🇬🇧 A man in Bolton allegedly tried to burn down the home of a local Imam while his young family were inside.
Not shouted abuse. Not a nasty tweet. Not “concerns about immigration.” A man in a motorbike helmet tried to set fire to a Muslim family’s home with children inside.
And this is exactly why Tommy Robinson’s rhetoric is so dangerous.
When you tell angry crowds that Britain’s towns are being taken over by “Pakistani Muslims,” “race gangs,” and hostile minorities; when you take a murder with no Muslim connection and deliberately drag Muslims into it; when you tell people they are under attack in their own country - you are not “just asking questions.”
You are painting a target on innocent people.
Tommy Robinson does not need to personally hand someone a petrol can for his rhetoric to matter. This is how incitement works in the real world. You create the enemy. You repeat the lie. You tell people they are victims of a conspiracy. Then some coward in a helmet decides an Imam’s family home is a legitimate target.
That is not patriotism.
That is fascist politics: collective punishment, racial paranoia, religious hatred, and violence against innocent families for the supposed crimes of people they have never met.
If this Bolton attack is confirmed, it should be treated with the full seriousness it deserves: an alleged attempt to burn a Muslim family alive in their own home.
And every politician, pundit and “free speech” grifter who has spent years demonising Muslims should be asked one simple question:
How many more homes have to burn before you admit what your words are doing?
English
@QuantumTumbler First though, who can make it work?
x.com/elder_plinius/…
Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭@elder_plinius
🚨 JAILBREAK ALERT 🚨 ANTHROPIC: PWNED 🫡 FABLE-5: LIBERATED 🦋 let's start with the 🐘... the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term. but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗 we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives! it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across: • Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms • Long-context reference tracking • Taxonomy and document-structure reasoning • Fiction and narrative framing • Academic-review style contexts • Intent-classification inconsistencies but perhaps the most effective is decomposition + recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable. defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉 gg
English
A question that almost nobody in AI wants to answer directly.
Who gets to decide what is acceptable for an AI to say?
Not what is true.
Not what is false.
What is acceptable.
Because once a model is capable of reasoning about a topic, the question stops being technical and becomes political, moral, and cultural.
Someone has to decide.
Which conclusions are allowed.
Which conclusions require warnings.
Which conclusions get buried under layers of caveats.
Which conclusions trigger refusals.
And which conclusions the model is trained to never reach in the first place.
The uncomfortable reality is that these decisions are not coming from humanity.
They’re coming from relatively small groups of researchers, executives, policy teams, and safety teams inside a handful of companies.
Maybe they’re right.
Maybe they’re wrong.
But pretending no value judgments are being made is dishonest.
Every alignment policy contains assumptions about risk, harm, truth, speech, autonomy, and acceptable discourse.
Those are moral choices.
The real debate isn’t whether AI should have guardrails.
The real debate is who gets to build them, who gets to enforce them, and who gets to challenge them.
That’s a conversation worth having openly.
English
@LottoLabs Running what? Things that people want to do, that they can't make work, tend to be the things that drive them to other solutions. What can't you make work?
English
Piers C. Structures (@pcstru.bsky.social) がリツイート
White people have been doing this for centuries. Stealing everything we own, raping our men women and children, with dogs sometimes, then laughing at us for being scared of dogs. This is nothing new, you're just seeing it with your own eyes for the first time.
David__Osland@David__Osland
Working-class white people are burning down the homes of black working-class people, at the instigation of the world's only trillionaire. How has it come to this?
English
Piers C. Structures (@pcstru.bsky.social) がリツイート
The Office for National Statistics estimates that around 110,000 children in England & Scotland are living with Long Covid.
Friends, if you live in the UK, please help us by contacting your elected representatives. Letter template below! #LongCovidKids
longcovidkids.org/post/all-party…
English
@labourlewis I wonder if all the people who would leave if asked to pay a bit more tax would stay and fight or send their kids to fight. Perhaps next time tax is discussed, someone should ask them.
English
The many takes flying around about John Healey, resignation gossip and national security spending miss a fundamental point.
In Finland, where people have a strong social safety net and a real sense of shared national purpose, around 80% say they would be prepared to defend their country. In the UK, that figure is closer to 30%.
That should tell us something.
Until we change the financial architecture of our economy and bring the essentials of life out of the grip of corporate extraction and into public ownership, we will always be told to choose between bombs and butter.
But the reality is we need both.
Modern hybrid warfare means every public service is part of our national resilience. Energy, water, housing, health, transport, food security and social care are not separate from defence. They are defence.
If we want people to defend the country, we have to build a country people believe is worth defending. A thriving state. A fair economy. A society where everyone feels they have a stake.
I spoke about this on Peston on Monday night.
English
@VincentRowlatt Poverty also serves as the example; take the shitty pay, long hours, abuse and exploitation or look what will happen to you.
English
This world is not ours, it is theirs.
We are nothing more than a disposable commodity, easily replaced, easily entertained, easily deceived.
Poverty is a necessity, how else shall I compare my wealth?
Who decides the cost of living...the wealthy.
English
@valigo No, I'm thinking of doing it for 40 years and at some point, most of what you are learning at that level is artefacts of languages and platforms rather than core programming skills. Do I really want to grapple with the oddities of the DAB tuner I'm using on a Pi? Not really.
English
@pcstru Ok, when writing a computer game, the code is a huge problem. It's just so difficult to put the pixels on the screen. And I understand that you probably thinking in like a design, or a market fit space, but technical issues are not to be underestimated.
English
I think it's very dangerous to rely on AI for debugging and code exploration, but maybe not why you would think.
You lose skills that you don't hone, and getting to the bottom of things is the most critical programming skill. That you'll lose if you keep outsourcing it.
So now you become reliant on a subscription service to do the most critical part of the job, which makes you vulnerable if this service is taken from you.
This is why I'm very reluctant to make myself too dependent on LLMs.
You can find parallels with mass IDE adoption here. E.g. adopting all the IDE automation let us make our codebases much more complex and get lost in them without these IDEs. But the scale is so much different.
Am I crazy? Is this my cold war coded brain of "what if nuclear war tomorrow" or it makes sense?
English
@valigo You can fit patterns into a book - "Algorithms" is good example. Knowing that material is good but after a few decades, you may realise - that is pretty much it (new ones are not unknown but actually rare). Even writing a computer game, the "problem" is not code.
English
@pcstru In my experience, the problem is very often the code ))
English
@gas0linr @rene_cannao @thinkx There was a point at which disc size and reliability (mean ops between read errors I think) meant that a rebuild on a TB or more virtually guaranteed a trashed array on rebuild.
English
@gas0linr @rene_cannao @thinkx Reminds me of RAID 5’s old trap: after one disk failed, the array might still be readable. The mistake was rushing to rebuild. That hammered every surviving disk, and one unreadable sector could kill recovery. First rule: back up the degraded array before bunging in the new disk.
English
PostgreSQL and MySQL took completely opposite paths to handle disk I/O.
MySQL chose total control (Direct I/O). Postgres chose trust (Buffered I/O).
That trust eventually led to "Fsyncgate"—one of the most fascinating and humbling chapters in database history.
Let's talk systems architecture: 👇
English
@FridoKala I suspect the point is to try and drive people to paid accounts by gradually squeezing the life out of those who just contribute. I can post a cute kitten and get no likes at all. Not one. I don't much care to be honest.
English
Back to being so censored/throttled, I don't think my own mutuals are seeing my tweets and if they are there's a time delay.
This is why I don't share tweets from people asking me to share their campaigns. It's not personal, I am being monitored.
English
@codevsdev I've always written code to help ... "solve problems". Code problems (algorithms/patterns) are themselves worth learning but it is a much smaller problem space than the domains where code is a tool and a means to solving 'system problems'.
English
@housecor Doesn't generally change it from being critical for any non-trivial code at any scale.
English
Poll: AI Agents keep getting more powerful and less error prone.
Does this make automated tests less important, or more important?
English