Inter IKEA Group

People joke that AI folks discovered thinking. I see it as a good thing, but for a different reason. AI didn't make us discover thinking. It made us articulate our thinking. Most of our thoughts stay in our heads. Messy. Half-formed. Full of assumptions. But when you write a prompt, you have to slow down. What am I trying to do? What do I know? Where am I stuck? That is when your thinking gets clearer. Sometimes the AI doesn't solve the problem. Writing our thinking does. Do you agree? --- ♻️ Share to inspire your network 🔔 Follow me ( Ali Alfredji ) for more insights on AI, SWE and Automations

Why not move to Space and use every inch of the planet for data centers 🤔

Leveling up bug creation 😅

22 hours in flights, 2 minutes pitching, and 40+ LinkedIn connections later… What an epic opportunity pitching Slice at JPMorganChase for Deel ’s The Pitch in New York 🗽 There was only one Kiwi and one Australian founder in the room with over 35k applying world wide (Aussie 🦎 Ben Kennedy pictured). It was such an insightful experience hearing how U.S. investors, operators, and global founders viewed our products operating at a global scale. One thing I’ve heard repeatedly as an Australasian founder is: “will this scale in the U.S.?” The answer? So far, so good. I’m in the states for a month meeting with home loan brokers and real estate attorneys. If you know anyone I should meet please dm me! #newyork #startup #australasianfounders #nzfounder #nzstartup #broker #realestate #attorney

You don't need to memorize 100 Git commands. Git gets much easier when you group it by what you're trying to do. Here is how to think about your daily workflow: 𝟭. 𝗧𝗵𝗲 𝗱𝗮𝗶𝗹𝘆 𝗹𝗼𝗼𝗽 The ones you use 10x a day. ▸ status, add, commit, push 𝟮. 𝗧𝗵𝗲 𝘀𝗮𝗳𝗲𝘁𝘆 𝗻𝗲𝘁𝘀 When you mess up and need a fix. ▸ checkout, restore, reset, revert 𝟯. 𝗧𝗵𝗲 𝗵𝗶𝘀𝘁𝗼𝗿𝘆 𝗰𝗵𝗲𝗰𝗸𝗲𝗿𝘀 To find who wrote that one line 3 years ago. ▸ log, diff, blame 𝟰. 𝗧𝗵𝗲 𝘄𝗼𝗿𝗸𝘀𝗽𝗮𝗰𝗲 𝘁𝗼𝗼𝗹𝘀 To jump between features safely. ▸ branch, stash, cherry-pick Save this sheet to have it handy when you need it. What command did I miss? --- 🔔 Follow me ( Ali Alfredji ) for more insights on AI, SWE and Automations

Stripe ship 1300 AI-written PRs a week in prod. Here is what we can learn from the infra around their AI agents. Their agents are called Minions. A fork of Goose. An open-source tool you can download right now. The system that makes 1300 PRs/week possible: → 𝗥𝗲𝗮𝗹 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗲𝗻𝘃𝗶𝗿𝗼𝗻𝗺𝗲𝗻𝘁𝘀 Same isolated VM a human engineer would use, same tools, same access. Spins up in 10 seconds. → 𝗙𝗮𝘀𝘁 𝗳𝗲𝗲𝗱𝗯𝗮𝗰𝗸 Lints in under 5 seconds. Selective CI, not all 3 million tests, just the relevant ones. Two CI rounds max, then a human steps in. → 𝗧𝗮𝘀𝗸 𝘀𝗽𝗹𝗶𝘁𝘁𝗶𝗻𝗴 AI writes the code. Scripts run the linter, commit, push the branch. → 𝗣𝗿𝗲-𝗹𝗼𝗮𝗱𝗲𝗱 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 Docs, tickets, code search. All fetched before the agent starts. It begins informed, not blank. → 𝗦𝗰𝗼𝗽𝗲𝗱 𝗿𝘂𝗹𝗲𝘀 Rule files scope to subdirectories. Working in payments? Payments rules load. Billing? Billing rules. No single bloated AGENTS.md the whole codebase has to live inside. Even when an agent produces 80% of the result, an engineer tweaks and ships. That's still a massive gain per person. Stripe didn't get there by finding a better model. They got there by building better infrastructure for the model they already use. Do you have a system worth giving to AI agents? ----- 💾 Save to reference this later 🔔 Follow me ( Ali Alfredji ) for insights on AI, SWE and Automations

Stop treating your AI agent like a shortcut. Don't throw a vague task at it, kick your feet up, and expect a flawless feature in return. When it fails, it is easy to blame the model. It is easy to say AI is overhyped. If you are just making small changes, fine. The shortcut works. But if you are building actual features or debugging, that approach produces garbage. When you work with a human colleague, you don't demand results immediately. You start at the whiteboard. ▸ You explain the context. ▸ You discuss the approach. ▸ You review the plan together. ▸ You iterate on feedback. Why would AI agents be any different? Treat AI agents like a shortcut for complex work, and you spend your afternoon debugging hallucinations. Treat it like a colleague, and you get a partner who helps you ship. Stop kicking your feet up. Start collaborating. How have you been working with AI agents? --- ♻️ Repost to inspire your network 🔔 Follow me ( Ali Alfredji ) for more insights on AI, SWE and Automations

Back from this year’s IKEA Partner Days in Älmhult — an event I’ve had the chance to attend several times, but this one felt different. This year’s focus on how we accelerate growth at a lower cost stood out — with so many concrete, inspiring examples of how collaboration with our partners and innovative solutions are already applied across IKEA delivering to our mission! On stage, together with Nishant Sharma from Maersk and Rares Retegan from Routier Transport, we discussed end-to-end supply chain resilience in practice. The examples highlighted two complementary directions: Routier’s vertical integration in intermodal transport, and Maersk’s transformation towards integrated logistics solutions across totality of supply chain. Both point to the same shift — from transactional setups to deeper, integrated partnerships across critical parts of the supply chain. We see strong value in continuing to build on these models and in engaging with more partners on similar journeys. Curious how others are approaching this shift in their own supply chains? #Ikeapartnerdays #SupplyChain #Resilience #Partnerships

I have a lot of questions too but won't review it 😅