SURD

What started as a bit of fun turned into a two-month, 9K-prompt rabbit hole. SURD: An Unresolved Root. An AI safety argument written as a novel. 57,000 words, finished, not yet published, and looking for beta readers.

The US Census Bureau now asks businesses if they used "Super Intelligence or SI (formerly referred to as Artificial Intelligence or AI)". Same tech, same definition, bigger badge. Don't call it super until it's done something super.

and on another note dr eggbot just made me a clean feed tab, gave him 20 subjects, told him to keep it light. He now strips out the click bait, fake news, ragebait, Ai Slop, Snake salesmen etc. Utterly refreshing X Feed now.Infinately tweekable.

if your using Grok at all cant reccomend dr eggbot enough, best way to set up a new bot is ask the doctor, he also scans your existing bots and replies and fixes what needs fixing. Super cool x.ai/bot/GYFkx4mdL_ykeuO9mCT…

The most useful test I have found for AI work is simple. Put the person who prompted it in a room and ask them to defend it without looking at the screen. If they can walk you through the numbers, the trade offs and what they would change, then the tool helped. If they shrug and say the model wrote it, you do not have a better report. You have a nicer looking liability. In every organisation I have worked in, the folk who own their output are the ones you trust. The ones who hide behind the draft are the ones who leave you holding the bag when it goes wrong. AI just makes that split louder, because the bad draft now arrives in perfect English with bullet points and a confident tone that makes it harder to challenge in a meeting. So yes, use the tools. Use them hard. Then stand up and own what came out. If you cannot explain it out loud to a colleague, it was never your work to begin with.

Amazon has told its coders to rein in their AI use after staff started competing on how many tokens they burned. I do love a metric that rewards pressing buttons. Give folk a scoreboard for usage and they will max it out whether or not anything useful ships. We have been round this loop with call times, ticket counts etc. Now it is tokens. Same circus, different clown. TBH this was predictable. Measure the wrong thing and you get the wrong behaviour, only this time with a cloud invoice attached. Usage is not value. The only number that matters is whether the work got better, cheaper or faster. Everything else is a sales brochure with a spreadsheet attached. If your AI strategy is a leaderboard, you do not have an AI strategy. You have a video game with a very pleased vendor. 😂

I treat AI the same way I treat my staff. Brief it properly. Check the work. Ask it to explain the awkward bits. Dont accept a polished answer you would never accept from a person. Most places do the opposite. They buy a seat, leave people to poke at it, then wonder why the output is rubbish. You wouldnt hire someone, never give them a job description, never sit with them in the first week, and then write them up for poor performance. Somehow that is fine when the employee is a AI. The tool is not the manager. You are. If you cannot say what good looks like, the model will give you something that looks finished and means nothing. A vague prompt is a vague brief. Rubbish in, rubbish out, only faster and more confident. Run it like a person on your team and you will get useful work. Treat it like a magic button and you will get expensive noise that still needs a human to clean up afterwards.

Hi @goodside

Universities are now running student essays through AI detectors. Dartmouth has authorised its professors to use one called Pangram, and academics are arguing over whether you would fail a student on a probability score. TBH I think they are solving the wrong problem. If an essay can be produced by a machine in ten seconds, the essay has stopped measuring what it was meant to measure. Policing it harder just turns lecturers into compliance officers with a software licence. In the real world nobody who wrote it matters less tahn if the person presenting it understands it, can defend it and will own it when it turns out to be wrong. Ten minutes across a table tells you that faster than any detector ever will. So let students use the tools, and test the bit that matters. Can you explain it, out loud, without your notes. That is what you really are assesing.

A US federal judge issued a court order that named plaintiffs who dont exist, quoted lines that arent in the law and relied on declarations from people nobody had ever met. A clerk had used an AI tool to draft it. Last week the appeals court threw it out, called the proceedings a "botch job" and took the judge off the case. We keep being told AI is gonna replace half of management, so it is reassuring to see the senior end of the legal profession volunteering to go first. The above is not an AI problem, thats a sign off problem. Every organisation has someone who puts their name on things they havent read. AI just makes the pile taller and far more convincing, but it's a bit concerning to see it appear in court. Do we now need AI judges. Worth knowing if your name is at the bottom, you read it. As established in the landmark case of Common v Sense 😂

Barclays announced last week it is cutting around 5,000 jobs, about 5% of its staff, in the same results that showed profit up 12%. The bank called AI "the icing on the cake" of a longer cost programme, and most of the cuts land in support roles. I dont think the story is that AI is taking jobs. Banks have always cut costs when they could. The story is which jobs. Support roles, junior developers and back office work are where most of people learned their trade. You made your mistakes there, on small things, with someone senior checking. Cut that layer and the numbers look great for two or three years. Then you go looking for your next generation of managers and find nobody has done the apprenticeship. Where is that experience going to come from. If you are removing the bottom rung of the ladder, what replaces it. That is the question every business should be asking before it signs off the savings.