PHP2Shell is one of the most serious web vulnerabilities of the year And most site owners still don't know if they're affected 2 flaws in PHP let anyone run code on your server without any login needed Around 70% of the web uses PHP, including almost every WordPress site GPT-5.6 Sol caught it back in August, and the fixes only went out on 21 October Exploit code is already circulating, so this one can't wait Here's a prompt that checks your site and tells you exactly what to do (check in comments)

THIS IS HUGE... Anthropic just announced Claude Duet Think Grok Bot or Dots, but built around Claude Here is what it can do: > runs on Claude Opus 5.5 > connects to 6,000+ tools and apps > carries over your Claude memory and skills > monitors your apps and sends proactive updates > builds dashboards, apps and projects in background > runs on its own cloud computer, even when you're offline The race to build the best personal AI agent is getting very interesting...

Elon just announced that Grok Bot will use whatever backend gives the best result for each task, Claude Opus 5.5 included... xAI builds its own models and still routes work to its competitors. That's how much routing matters... You can do the same in your own harness, here's how: > list the tasks you run most: planning, writing, coding, research, images, small yes/no decisions > assign one model to each: your best model for the work people read, a cheap one for sorting and checking, specialists for images and audio > write that list into your instructions file (CLAUDE.md, AGENTS.md, or your agent's custom instructions) as a routing table, so it's read every session and you never pick a model by hand > let the plan decide: once you approve a plan, each sub-agent runs on the model that matches its task > log which model ran each step, what it cost and how it went, so you can see where a cheaper model is good enough Once you find which models are best, the speed, quality and cost all move in your favor Full routing table + CLAUDE.md section in the image 👇

GPT 6.1 SOL is now around 20% faster than it was around DevDay The only alternative to achieve faster speed is sub-agents, but they come with a cost: - delayed time-to-first-token (cause of seconds spent routing) - higher system complexity (requires orchestration) - lower cost efficiency So overall this is a great improvement

If you want to grow on X, save this prompt: --------------------------------------------- You are my content strategist. Below is an export of my posts from the last 90 days with their metrics (views, likes, replies, reposts, bookmarks). Analyze them and produce four deliverables. Before you start: calculate my median views and engagement. A post is a "winner" if it performs well above my median, and a "flop" if it's well below. Judge everything against my own baseline, never against general benchmarks. ANALYTICS REPORT My baseline numbers (median views, likes, bookmarks) My top 10 posts and bottom 10 posts, with the metrics The patterns that separate winners from flops: topics, formats, length, hooks, use of media, day and time posted What gets saves vs what gets likes (they're usually different) 3 things I should stop doing, with the posts that prove it Only state patterns backed by at least 3 posts. If the data is too thin to tell, say so. WEEKLY CADENCE How many posts per week, and on which days and times, based on when my winners went out A 7-day plan: for each slot, the topic, the format, and which winning post it's modeled on A mix that keeps my best-performing format as the backbone, with room to test one new thing per week VOICE.MD Write a markdown file another AI can use to write exactly like me: Tone and personality in 3-5 lines Sentence style: length, punctuation, capitalization, formatting habits Words and phrases I use often, and ones I never use How I open posts and how I end them 3 short examples of my writing at its best, quoted from my posts A "do not" list of things that would sound unlike me HOOK LIBRARY Every opening line from my top 20 posts, ranked by performance Group them into hook types (question, bold claim, number, story, contrarian...) For each type: why it works for my audience, and 3 new fill-in-the-blank templates based on it Use only what's in my data. Don't invent metrics or quote posts I didn't write. [PASTE YOUR POSTS AND METRICS HERE]

Everyone is posting graphs of how many people use AI and how early we are (mostly on LinkedIn) But I don't see many posting graphs specific to businesses. So I went through some of the surveys and the numbers here look a lot different: - 32% of companies decided to not buy a software and built the thing internally with coding agents instead - 28% now put more than a tenth of their entire IT budget into AI - 37% report that AI has contributed positively to their organizations’ EBIT - among companies over $1B, 40% have agents scaled across the enterprise, up from 27% in a single year - 31% of those have coding agents running in production - of the ones actually getting value, 72% fundamentally redesigned their workflows around it So, if you're a smaller company you're still early if you start now. But as for larger ones, your competitors are already implementing it

AI can drastically grow your company, but you need to fix the layer underneath first Cause automating a process that doesn't work is just making that same process faster The order that actually works: question why each step exists, delete what you can't justify, simplify what's left, use data, then automate it A lot of founders skip these steps and just automate randomly: - automating a report that is not being used - automating chat replies, without training it on chats that closed deals - sloppy content, that is again not based on data and your work You can build as many automations of these as you want, it will not work in your business Fix the layer first, build something that works, collect data and make it compound instead of just going faster

How to build an AI system your team will actually use: - diagnose before you build - pick one process, not all of them: weekly at minimum, same pattern every time, a mistake you can survive - extract the knowledge, don't write it: point it at real conversations - write the guide: your employees will forget things, new employees will come.. - log seven fields from run one: timestamp, process, inputs, model, output, result, cost - don't give full freedom from the start

Instead of hiring 3 people, hire 1 good one and build the AI layer that does the repetitive half of their week Wrote the full method

Took the evening off in Dubai AI agents handled the business, checked once, nothing needed me this is the whole point of what I build https://t.co/4QXV4O8F9b