Exemples de lead magnets LinkedIn en project management

4 jours pour te lancer sur Claude Code. Gratuit. 1 session d'une heure, du lundi au jeudi. J'ai passé +200h sur Claude Code et accompagné +20 indépendants et entreprises. La semaine prochaine, je te donne les clés, les concepts essentiels pour te lancer. Du lundi 4 au jeudi 7 mai, 4 sessions live de 7h30 à 8h30. Format simple : 20 minutes de présentation, 40 minutes de Q&A. À la fin de la semaine, tu seras à l'aise pour démarrer, tu sauras pour quoi t'en servir, et tu connaîtras les grandes étapes pour construire ton système IA. Le programme : Jour 1 - Lundi : Passer le pas Prendre Claude Code en main dans l'interface Desktop, et les fondamentaux pour bien démarrer (dossiers, plan mode, premiers réflexes...) Jour 2 - Mardi : L'importance du contexte L'IA n'est pas bête, elle manque juste de contexte. Comment lui en donner pour qu'elle bosse comme toi. Jour 3 - Mercredi : Les skills Transformer tes process récurrents en commandes que tu relances d'un mot. Jour 4 - Jeudi : Les agents Pourquoi un seul Claude ne suffit pas, et comment construire ton équipe IA spécialisée. Tu ne peux pas venir un jour ? Pas grave. J'envoie les replays chaque jour à tous les inscrits. Pour participer : 👍 Like ce poste ✍ Commente ce que tu veux 🤝 Envoie moi une demande de connexion Je t'envoie l'invitation en message privé. Back to work ! ✌

karpathy connected obsidian to claude code and something clicked his setup: obsidian vault as a knowledge base. claude code reads it as context. every note, every link, every tag becomes part of what the agent knows but here's the part most people skip the vault is not just notes. it's structured context. each file has frontmatter (metadata), backlinks (relationships), and tags (categories). claude code reads all of it this means your agent doesn't just know what you wrote. it knows how your ideas connect. which projects reference which resources. which clients have which problems. which patterns keep showing up i've been running this for months. my second brain has: - client folders with README, communication log, technical docs, progress - a CLAUDE.md that tells claude who i am, how i think, how to work with me - 60+ skills that extend what claude can do - persistent memory across sessions the obsidian founder (kepano) even released his own claude code skills. 19.9K stars. they teach claude how to work with markdown, bases, json canvas, and the obsidian CLI this is not a note-taking app anymore. it's an operating system for your business when you connect it to an AI agent comment 'brain' if you want the setup guide

If you can’t personalize at scale, LinkedIn wins. If you can’t afford SDR bandwidth, email wins. If you need meetings NOW, calls win: Here’s the reality of outbound in 2026: 1/ Cold Email → Cheap, scalable, perfect for testing messaging. But reply quality is mixed, and deliverability is always a fight. 2/ LinkedIn DMs → Relationship-driven, intent-based, and fast to get replies. But you’re capped by daily limits and the need for high-quality content. 3/ Cold Calling → The fastest way to qualify big deals. Instant feedback and high trust. But it’s expensive, tough to scale, and reps burn out fast. Don’t crown a single winner. Build a system where each channel covers the gaps of the others. Want me to help you rebuild your sales fundamentals with The Modern Seller proven system? DM me "Modern" and let's see if we're a fit (allowing only 3 new projects per month) Or grab the (FREE) 10 cold email frameworks that books me +30 meetings per month: https://lnkd.in/e4QY3TPW

Amazon cuts another 16,000 workers. Just the start of the AI takeover... In 2026, over 83M jobs are predicted to be replaced by A.I If you work any of these 6 jobs, you might be in trouble… 1- Finance & Banking Wall Street expects to cut 200,000 jobs with AI in the next 3–5 years. A Bloomberg survey found major banks plan to shrink workforces by up to 10% as AI takes over analysis, trading, and compliance.​ 2- Human Resources: AI is automating up to 60% of routine HR tasks: payroll, screening, scheduling, performance tracking. PwC reports that 75% of CEOs expect generative AI to radically change business within 3 years.​ 3- Project Management: World Economic Forum research shows 50–60% of junior tasks like drafting reports, rescheduling, and updates, can already be automated by AI. White-collar jobs are restructuring fast, and strategy roles won’t be spared.​ 4- Accounting & Auditing: Accounting is third for predicted job loss. AI can automate reconciliation, auditing, and tax prep. Goldman Sachs forecasts 18% of all work tasks automated globally by AI within five years.​ 5- Legal Services: AI now reviews contracts and research at scale. Up to 30% of tasks in legal services are automatable. McKinsey predicts 60% of legal work will be substantially altered by AI tools by 2030.​ 6- Tech & Analysis: 77,999 tech jobs cut in the first half of 2025 were directly linked to AI. Half of entry-level tech, analyst, and reporting roles could disappear in five years as algorithms do the heavy lifting.​ I don’t want to be the bearer of bad news… But if you’re in one of these industries you only have 2 options: Ignore what’s going on and pray you won’t get replaced. Start leveraging the technology & “A.I proofing” yourself. If you picked option 2… You have to realize… This technology isn’t your enemy, it’s an opportunity. All it takes is putting in some time to stay ahead of the curve. If I was back working in my corporate job, here’s exactly what I’d do to protect myself: #1: Leverage A.I at Work Use Superhuman.AI to manage your inbox. Use ChatGPT to write your e-mails and presentations. Use Motion.AI to organise your calendar. If you’re good at using A.I, your boss will find it WAY harder to replace you. #2: Diversify Income If ALL jobs end up getting replaced… It won’t matter how good you are at using A.I in your job, you could get laid off. So I’d start using it to build a 2nd stream of income right now. Here’s how… As a corporate worker, you have a HUGE advantage right now. These tech kids understand coding, but don’t know real-world applications. You take existing tools, understand the problem it solves, then help a business owner use it. I’ve made $1m+ in the last 2 years doing exactly that. I don’t have time to go into full detail in this post… So if you want a free guide where I break down the exact business model… DM me playbook “PLAYBOOK” & I’ll send it over, no strings attached!

Most people use Claude like a search engine. Type a question. Get a response. Close the tab. Start over tomorrow. That's not an AI employee. That's Google with better grammar. I spent the last 6 months setting up Claude Co-work to actually run parts of my business. Not answer questions. Run tasks. On its own. While I sleep. Here's the difference: A chatbot answers questions. An AI employee does work. It knows my business. It knows my voice. It has access to my tools. And it runs workflows without me standing over its shoulder. There are 3 layers that make this work: Layer 1: The Role Skills, a claude.md file, and project memory. This is how Claude learns your business, your processes, your tone. Without this, you get generic outputs you spend 20 minutes editing every time. Layer 2: The Tools Connectors to Gmail, Slack, your CRM, your project management tool. Without these, Claude is stuck in a text box with no way to actually do anything. Layer 3: The Triggers Slash commands and scheduled tasks. This is what puts it to work automatically. My morning briefing runs every day before I even open my laptop. I built a content workflow that pulls my YouTube transcripts, writes platform-specific posts in my voice, formats them, posts to Slack for my team to review, and reports back with confirmation. 6 steps. I touch zero of them. This took me from "Claude is cool" to "Claude is doing a job." I recorded the entire setup. Every layer. Every step. How I actually built it. Comment "WORKFLOW" and I'll send you the full Setup instantly.

30 900 étoiles GitHub pour diviser tes coûts IA par 71. IA : le truc qui rend Claude Code vraiment utilisable vient de passer 30 900 étoiles sur GitHub. Ça s'appelle Graphify. Et même si tu touches pas une ligne de code, tu devrais t'y arrêter deux minutes. Le problème, tu l'as forcément vécu. Tu accumules des fichiers pendant des mois. Du code, des PDF, des notes Google Docs, des captures, des replays de réu. Tu sais qu'il y a de l'or là-dedans. Sauf que le jour où tu veux t'en servir, tu passes une heure à fouiller ton dossier comme un archéologue bourré. Ou pire : tu demandes à une IA « lis-moi ces 50 fichiers et réponds ». Et là, le compteur s'affole. Chaque question = tout relire. Pour ceux qui bossent sur Claude Code ou Cursor, c'est encore plus violent. Tu reprends un projet trois semaines plus tard, l'assistant doit se réorienter à l'aveugle dans ta base qui a doublé de volume entre-temps. Le /resume fait le taf pour le contexte général, mais la structure profonde du code ? Oubliée. CE QUE GRAPHIFY CHANGE Au lieu de faire relire toute la bibliothèque à l'IA à chaque prompt, Graphify lit UNE fois. Il construit une carte interactive : les concepts, leurs liens, les dépendances. Ensuite, l'assistant consulte la carte au lieu de se retaper les 50 fichiers. Résultat annoncé : jusqu'à 71 fois moins de tokens par requête sur un corpus mixte de 52 fichiers. Pour ceux qui vivent pas avec le compteur Anthropic ouvert en permanence : une requête à 1 euro tombe à 1,5 centime. Et elle répond plus vite. LES 3 TRUCS QUI ME PLAISENT Multimodal pour de vrai. Code, PDF, images, vidéos YouTube, audios transcrits EN LOCAL sur ta machine. Rien ne part dans le cloud côté audio. Dans une époque où chaque outil rêve de siphonner tes données, c'est pas un détail. Chaque lien de la carte est étiqueté « trouvé » ou « deviné », avec un score de confiance. L'outil te dit noir sur blanc ce qui vient des sources et ce qui est une inférence. Dans un monde où les IA hallucinent en souriant, cette honnêteté est rare. Mise à jour incrémentale. Tu modifies 3 fichiers sur 50, Graphify ne retraite que ces 3-là. Pas de rebuild complet toutes les 10 minutes. POUR QUI Quiconque gère un corpus qui grossit, base documentaire, dossiers clients, doc technique, projet de code, et qui veut l'interroger avec une IA sans se ruiner en tokens ni poireauter 30 secondes pendant que l'assistant se réoriente. Pendant que tout le monde court après le prochain modèle à 2 000 milliards de paramètres, les vraies ruptures de productivité se jouent sur la PLOMBERIE. Ce qu'on donne à lire à l'IA, comment, et combien de fois. Le cerveau le plus rapide du monde reste con s'il doit relire la même bibliothèque à chaque question. Lien en commentaire. ____ Eudonia eudonia . fr Structurez l’analyse de vos accompagnements en quelques minutes

You've heard that SKILLS are one of Claude's superpowers. But you're still writing the same prompt every time.  --- Each session, you type the same instructions. "Read this article. Pull the key insights. Give me something I can use." Claude gives you a summary. It varies. Different structure. Different depth. That's because Claude starts fresh and doesn't remember what worked last time. Without a skill, you're the middleware. Run step one. Copy the output. Paste it into step two. Check the result. Feed it into step three. Multi-step workflows are multi-step chores. A skill is a set of tasks written in plain English. Step-by-step instructions with the context baked in. What to do, in what order, what must not happen. The skill lives in a folder in your Claude project. Claude reads it when you need it, runs every step, and gives you the final result. ✅ Intermediate steps just happen. ✅ Outputs are consistent. ✅ Workflows are repeatable. Think of skills as tools; things Claude can reach for when the job matches. Build enough skills and you can wire them into agents that handle entire workflows without prompting. Skills are building blocks that help you achive more.  --- Here's a real example. Insight Harvester Skill. I give Claude a URL of an article I found. Claude reads the article, summarises key insights, asks me three questions to get my thoughts, saves my take into an "insight" file, and generates three content ideas shaped by my position. One URL in. One POV documented. Three content ideas to work up. Less than two minutes. Before this, I'd read an article, think "that's interesting," and forget it by Thursday. Or how about my Meeting Prep Skill? I give Claude a name, a LinkedIn URL, and the meeting purpose. Claude researches the person, finds common ground, drafts talking points, three opening questions, and saves a one-page brief. Ninety seconds. And I walk into meetings prepped and ready to go. Building each skill took less than 20 minutes. One folder. One markdown file. No code required to get started. And the sky's the limit with how far you want to take things. --- The carousel walks through every step. Comment SKILL and I'll send you the starter kit; a how-to guide including both skill templates. ♻️ Repost if someone in your network is still prompting from scratch every session. 🔔 Follow Tom Jones for more on AI and building systems that work.

I switched to Claude Opus 4.6 for my hardest work. And honestly... I don't think I'm going back. Here's everything it can actually do: THE MODEL 1. Claude Opus 4.6 ↳ Think of it as the one you bring out when things get complicated ↳ Long projects, deep thinking, tasks that would make your brain hurt ↳ Basically... the big gun I'm sharing all the PROMPT PATTERNS you need to get 100x more out of it.  (Stick around till the end, it's worth it) THE FEATURES 1. Adaptive Thinking ↳ It figures out how hard to think based on what you give it ↳ Like having a colleague who actually reads the brief before responding 2. 1M Token Context ↳ You can throw an entire project at it and it won't forget a thing ↳ Entire folders, documents, files... all at once 3. Context Compaction ↳ Long conversation? It quietly summarises the old stuff so nothing gets lost ↳ No more "wait, what were we talking about again" 4. Agent Teams ↳ Multiple Claudes working together on the same problem ↳ Yes, you read that right 5. Web Search ↳ Live search with sources attached ↳ Great for when you need info from, you know... today 6. Research ↳ Deep web research that actually synthesises what it finds ↳ Not just links... real answers 7. Claude Code ↳ Builds, debugs, and refactors entire repos on its own ↳ Developers, this one's for you (Max plan, $30/mo) 8. 128K Output ↳ One response can be an entire report, doc, or codebase ↳ No more "continue" buttons So yeah. I use Opus 4.6 when things get complex. Sonnet 4.5 for the everyday stuff. Between the two, I've genuinely stopped switching between 10 different tools. Want the high-res PDF + PROMPT PATTERNS? 1. Like this post 2. Comment "OPUS" 3. Connect with me and I'll send it straight to your DM