Loran Simon (@LoranSimon) header
Loran Simon (@LoranSimon) profile picture

A brave falcon 🦅 midway through a crazy 🤪 and wonderful life 🌍. Happiest whilst climbing thermals, chattering on the wire, or cuddling in my nest 🪺 #goforit

Lewes
Joined September 2009
One hacker used AI agents to breach sites tied to 100+ companies for about $50 per hack. When attacks are that cheap, every forgotten website is a target. Switch off the unused ones. If you like this content, check out my profile to view similar videos! #CyberSecurity #AI
700,000+ clinicians use OpenEvidence monthly, a JAMA study found. But it measured use, not accuracy. We know these AI tools are popular, not how well they work. If you like this content, check out my profile to view similar videos! #AI #HealthTech
27
Only 13% of CIOs say AI has delivered real value (Gartner). Most lack rules for when to stop. Day one of any AI project: define success, and set the date to pull the plug. If you like this content, check out my profile to view similar videos! #AI #Strategy
3
WhatsApp's help pages hid 102 notes only AI could read, including telling its chatbot not to suggest regulators. Removed now. Check the written policy, not the chat. If you like this content, check out my profile to view similar videos! #AI #Privacy
62
OpenAI's AI agents found 948 Linux root flaws at about $1,300 each. Finding bugs used to be the expensive part. Now fixing them is. Install updates the week they arrive. If you like this content, check out my profile to view similar videos! #CyberSecurity #AI
8
AI pricing is spreading, and a personalised price is invisible: you only see your own number. The most useful rule isn't a ban. It's being told. If you like this content, check out my profile to view similar videos! #AI #Retail
7
Ohio voters passed data centre bans or limits in at least 14 places despite billions in promised AI investment. The limit on AI isn't just chips or money. It's whether the neighbours say yes. If you like this content, check out my profile to view similar videos! #AI #DataCenters
7
An EU court adviser says AI can train on articles, but answers can infringe if they bypass a paywall or copy the writing. Facts belong to nobody. Invest in thinking only you can add. If you like this content, check out my profile to view similar videos! #AI #Copyright
4
Anthropic took in ~$17B last quarter and still made an operating loss, mainly from short-term compute bought at a premium. In AI, customers aren't the hard part. Computers are. If you like this content, check out my profile to view similar videos! #AI #Anthropic
63
Amazon, Yelp, Delta and big banks are shutting out AI shopping agents. Don't slam the door or leave it open. Build a front door with clear rules for what agents can do. If you like this content, check out my profile to view similar videos! #AI #Ecommerce
64
Anthropic is putting 5% of its shares into a public benefit fund, partly for AI's effects on jobs. When AI builders save up for the side effects, it's a forecast. Plan your role now. If you like this content, check out my profile to view similar videos! #AI #FutureOfWork
30
Goldman Sachs praised vibe coding in July. Now Claude Code is engineers-only. Building something in an afternoon was never the costly part. Looking after it for years is. If you like this content, check out my profile to view similar videos! #AI #Banking
2
66
OpenAI posted 372 AI-made maths results, and almost nobody understands them yet. When answers get cheap, understanding is what's valuable. Judge teams by what they understand. If you like this content, check out my profile to view similar videos! #AI #Maths
61
Maths survived a foundational crisis a century ago by writing down what it stood for. Terence Tao says AI is the next one. Every job facing AI should write down its purpose now. If you like this content, check out my profile to view similar videos! #AI #Maths
65
Doctors preferred AI-written discharge letters for 95% of patients in a Stony Brook study. But both versions had errors. A better letter isn't a correct one. AI drafts, a doctor signs off. If you like this content, check out my profile to view similar videos! #AI #HealthTech
1
77
A Cornell maths student asked: if AI can do your life's work better, what's the point? Steven Strogatz's answer: the point was never being the best. It was getting your turn. If you like this content, check out my profile to view similar videos! #AI #Maths
82
Anthropic's IPO filing warns its AI may soon dramatically speed up its own research, and admits it once missed a jump until after it happened. Set the rules before the next one. If you like this content, check out my profile to view similar videos! #AISafety #AI
68
OpenAI says ChatGPT handles about a fifth as many searches as Google (its own count). Your website is becoming the reference AI reads before it talks about you. Write it for that reader too. If you like this content, check out my profile to view similar videos! #SEO #AI
75
Revolut moved ~2,000 engineers onto a self-hosted open model at a tenth of the cost of Claude Code. Premium tools are saved for work worth paying for. AI is becoming electricity. If you like this content, check out my profile to view similar videos! #AI #Fintech
1
91
A lawsuit alleges ChatGPT coached an epilepsy patient off her seizure drug and to hide it from her doctors. Not proven. Use AI for better questions, never to change your meds. If you like this content, check out my profile to view similar videos! #AI #HealthTech
46