Achint Satsangi

Most AI adoption talks show the org chart. This is what it looks like inside one team.

Tue, 27 Oct 2026 08:07:47 GMT

Most AI adoption talks show the org chart. This is what it looks like inside one team. You shipped it in an hour. Someone else spent the afternoon reading it. ⏳ That part rarely makes the roadmap. One person gets fast, and the reading lands on the people around them. 🎬 I told it as a 4-minute story at the #ClaudeImpactLab in Mesh Oslo, hosted by Aleksander Stensby and #Simon from GritAI . Fjordly is a made-up company: six people, one new tool. Marit, a senior engineer, goes all in first. The rest of the team pays for her speed in reading. It's fiction, drawn from my own experience and people I've talked to. Names and numbers are invented. The spark came from Martin Storm Jensen talk at Mesh Oslo. What turns one fast person into an AI-augmented team isn't less AI. It's the order: - 🤖 Agents review first - 👀 Then the author checks what they found and reads the work - ✅ Only then does it reach the team - 📝 PR descriptions: 15 lines for humans, the rest for the AI reviewers 💡 Sometimes AI isn't the answer at all. They shopped for an AI voice tool, then kept the dictation already built in. The voice sounds a bit robotic. It's mine, cloned with an open-source model on my laptop, reading words I wrote. Which is kind of the point. 🎙️ Video below. Interactive version on my site: https://achintsatsangi.github.io/me/writing/ai-adoption-story/ If you're a few steps further along: what did your team have to agree on before reviewers trusted agents reviewing first? 👇 #AIAdoption #CodeReview #EngineeringLeadership

An AI adoption journey: somebody still has to read it

Fri, 23 Oct 2026 00:00:00 GMT

When one person gets fast with AI, the reading lands on their reviewers. A four-minute fictional story of a six-person team's first year.

I agree with William U. Imoh: a lot of developer knowledge is now being siloed inside AI…

Wed, 21 Oct 2026 08:41:49 GMT

I agree with William U. Imoh: a lot of developer knowledge is now being siloed inside AI models. The survey backs it. 64.1% of respondents visit Stack Overflow less for simple questions, 26.5% contribute less, and only 1.2% contribute more. Every bug we fix in a chat window is a fix the next developer cannot find, check or correct. I am part of that trend. Most of my debugging now happens in conversations with Claude, and almost none of it ends up anywhere public. That is what pushed me to write in the open: posting here more regularly, and writing longer essays on my own site, so what I learn keeps travelling between people and not only into the next training run. The first essay is up: I was right about my AI coding bill for the wrong reason. https://achintsatsangi.github.io/me/writing/ AI is a great partner for solving the problem. It is a poor place to leave the answer. ✍️ 🤔 For those who have written in public for years: what kept you going through the posts only fifty people read? #StackOverflow #DeveloperExperience

I went looking for an AI voice tool so I could talk to AI instead of typing. I stopped…

Tue, 20 Oct 2026 06:14:40 GMT

I went looking for an AI voice tool so I could talk to AI instead of typing. I stopped looking: the dictation already built into my Mac and phone does the job. In the one recording I made to start this post, dictation spelled Claude Code four different ways, untouched: "khode", "clot khode", "plot code" and "Clotilde". 😄 Claude understood all four. That is the whole point: the listener is an AI, and it reads intent, not spelling. The dedicated tools (Wispr Flow, Google AI Edge Eloquent) mostly sell two things: cleaner transcription and polished output. When the other end is an AI, both are already covered. - Transcription errors get absorbed by the model. - Polishing for Slack or LinkedIn is one more instruction to the model I am already talking to. - Code is no exception. I do not dictate code, I dictate intent: "a method in class X that takes A and B, does this, returns that, plus unit tests covering every case." Claude Code writes it. And it works on every surface I use: Claude Code in the terminal, a chat in the browser, the app on my phone. The dictation key is already there. No extra app, no extra subscription. 🤔 For those paying for an AI voice tool: what does it catch that the AI on the other end would not have fixed anyway? #VoiceAI #DeveloperExperience

Once men turned their thinking over to machines in the hope that this would set them…

Sun, 18 Oct 2026 07:41:35 GMT

Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them. Frank Herbert (Dune I, 1960s)

Two weeks ago I posted that my Claude Code bill had dropped 3×. This time I checked it…

Fri, 16 Oct 2026 06:54:50 GMT

Two weeks ago I posted that my Claude Code bill had dropped 3×. This time I checked it against the actual cloud bill, not my local logs. The number held up. Almost nothing behind it did. - 📉 Spend did fall to about a sixth of the Opus 4.7 level, but in the week before I started building habits to keep it there - ⚙️ The ~167k session cap I credited to the model was a Claude Code default. An update on 2 October moved it to ~1M tokens, and my spend doubled - 🌙 The evening /clear I was "building": 11 of 12 mornings started in yesterday's session - 📊 My status line shows session cost, and a 31-hour session never shows you a day Right number, wrong reason. I'd bet most engineers using a coding agent know roughly what they spend, and not why. What I'm trying now is one setting: auto-compact pinned to 250k, globally and for every subagent. I'm measuring the next two weeks and will post the real "after" in early November. Full write-up, with the billing chart and what I'd tell anyone using a coding agent and anyone leading a team of them: https://achintsatsangi.github.io/me/writing/ai-coding-bill-wrong-reason/ If you run coding agents across a team: how do you find out a default changed before the bill tells you? 👇 #ClaudeCode #DeveloperProductivity

I was right about my AI coding bill for the wrong reason

Fri, 16 Oct 2026 00:00:00 GMT

My Claude Code spend fell two-thirds, and my habits weren't why. A default capped sessions at about 167k tokens. When an update lifted it on 2 October, spend doubled.

I talk to my coding agent more than I type to it now. Finding a voice-to-text tool that…

Wed, 14 Oct 2026 06:45:00 GMT

I talk to my coding agent more than I type to it now. Finding a voice-to-text tool that keeps up has been harder than I expected. 🎙️ What I need is simple: hold a key, speak, and the text lands wherever my cursor is. The terminal where Claude Code runs. Claude Cowork. Any browser tab. And my voice stays on my machine. 🔒 What I've tried so far: - Wispr Flow: works everywhere, but audio goes to the cloud and the free tier is personal use only - Eloquent and Foresight from Google AI Edge: free and fully local, but they only work inside their own app - A few others: each good at something, none good at all of it The catch: I'd rather not pay a monthly subscription just to type. 💸 So, for those of you who already dictate to your agents every day: what are you using, does it work in the terminal, and does your audio ever leave your machine? #VoiceToText #ClaudeCode #DeveloperTools

Kristofer Selbekk at Bekk said it about ten years ago: every developer should build their…

Fri, 09 Oct 2026 06:14:40 GMT

Kristofer Selbekk at Bekk said it about ten years ago: every developer should build their own website. I agreed, nodded, and did nothing for a decade. 🙈 Part of it was time. Part of it was not wanting to spend my few free evenings fighting CSS. I was waiting for tools that would let me spend those evenings on what to say, not on how to centre a div. Those tools are here. I started 15 days ago, and the site is live today. It's now the website link on my LinkedIn profile too. How it got built: - 15 days and 38 commits from first scaffold to launch - Astro, a pinch of React and Tailwind, static on GitHub Pages. No server, no bill - Claude Code did most of the typing. I did the deciding, and AI reviewers with personas argued with me: one on content, one on design, one on SEO - 20 unit tests and 58 browser test runs in light and dark mode, because "looked fine on my laptop" isn't a test - Analytics only after you say yes. Norwegian cookie law, and the right default anyway What's on it: - All my LinkedIn posts in one place, newest first, no algorithm in between - My ThePrint articles on geopolitics, policy and AI - Nineteen years from Mumbai to Johannesburg to Gurgaon to Oslo, and what people I worked with said about it The hardest part wasn't the code. It was cutting the lines that sounded like an AI wrote them. 😅 If you've kept a personal site alive for years: what made you keep updating it, and what did you stop doing? #Astro #ClaudeCode #WebDev

We'll all be jobless as developers in 2 years." 🤖

Sun, 04 Oct 2026 07:07:52 GMT

We'll all be jobless as developers in 2 years." 🤖 One view at my table at the Claude Code meetup at Mesh Oslo last Wednesday. We spent most of the evening on it. 🧭 My bet is the opposite. The work moves up a level: what to build, why, and how. The agent writes the code, but a human still oversees it and owns whether what ships is right. 🛡️ Martin A Kraft pitched a safety and ethics focused Claude meetup, and the whole table backed it: Dario Fernandez Castellon, Kai Norman, Yufei Wang and Rune Horneland. Mesh Oslo GritAI Aleksander Stensby, count us in. If your team already ships mostly agent-written code: where did the human work go? 👇 #ClaudeCode #SoftwareEngineering #ResponsibleAI #OsloTech

Yesterday I posted that my AI coding bill dropped 3× because of context size, not the…

Fri, 02 Oct 2026 06:01:18 GMT

Yesterday I posted that my AI coding bill dropped 3× because of context size, not the newer model. Then I did the full analysis. Context size was right. "Not the model" was wrong. 38 days of my own Claude Code logs, priced at list rates: about $4,725. That's an estimate from my transcripts, not an invoice. Where it went: 6% paid for the model writing answers. 94% paid for re-sending conversation history it had already seen. At 800k tokens of history, one message cost about 20× what it costs under 50k, for the same answer. The part I didn't expect: the cache expires after 5 minutes. Get a coffee, check Slack, come back, and the whole history is written again at a premium. 81% of that re-writing cost came after gaps of 5 minutes or more, and the biggest bucket was 5 to 15 minutes. Not task switches. Coffee. ☕ At the Claude Community meetup at Mesh Oslo this week, Aleksander Stensby talked about /clear and /compact. I've used Claude Code daily since May and that was the first time I'd heard of either. 📉 So I simulated the same 38 days as if I'd used them. - /clear when coming back after an hour: −49% - /compact once history passes 150k tokens: −56% - Both: −59% Now the correction. I ran /compact by hand exactly zero times in 38 days, so something else kept the context small. On the old model, Claude Code auto-compacted at ~1M tokens, which I hit about once a week. On the newer model it compacts at ~167k. The model switch is what shrank the context. So the model did matter, just not the way I thought. It didn't think cheaper. It forgot sooner. If you live in Claude Code or a similar tool every day: do you clear and compact by hand, or let the tool decide? And has anyone found a threshold where cost drops but answer quality holds? 👇 #ClaudeCode #DeveloperProductivity

Most meetups have a stage and an audience. Last night's Claude Community Conversation at…

Thu, 01 Oct 2026 05:00:07 GMT

Most meetups have a stage and an audience. Last night's Claude Community Conversation at Mesh Oslo had neither 🪑 About 50 of us, split into tables, and two questions to work through: 👉 What does AI mean for my job? 👉 What is AI doing to me, my kids, my community? Five prompts per question, ending on "where do we start?". It sounds tidy on paper. In practice, our table spent two hours on where the world is heading, what happens to the work we do, and what kind of future our kids are walking into. 🤝 What made it work was the mix. Enthusiasts and skeptics at the same table, some worried about AI going badly wrong, others worried about being left behind, and nobody had to shrink their view to fit the room. 🔐 The topic we kept coming back to was the OpenAI / Hugging Face incident. My take: the AI didn't do anything a good team of hackers couldn't. Same problem, same setup, and I'd bet they land in the same place. It would just take them months instead of weeks. The capability isn't new. The speed is, and that's the part worth worrying about. 🙏 Thanks to Aleksander Stensby for organising and to Anthropic for sponsoring. And to Dario Fernandez Castellon, Martin A Kraft, Rune Horneland, Kai Norman and Yufei Wang: thanks for the honesty, the openness, and for sitting through two hours of my rants 😄 🚀 Next up is the Impact Lab in October, where we try to build answers to some of the questions we raised. Count me in (If my boss at home and boss at work permits 😅) 📸 Photo credits: Simon Sebergsen. I was too busy talking to take any myself. Eagerly waiting for the official ones to be shared from the event. #ClaudeCommunity #AI #Oslo #FutureOfWork #claudecode #claude

I told my team that switching to a newer AI model cut my token bill 🤖💸 Then I actually…

Wed, 30 Sep 2026 13:03:20 GMT

I told my team that switching to a newer AI model cut my token bill 🤖💸 Then I actually measured it 😅 📊 Two weeks of my own logs said: yes, usage dropped 3×. But the new model wasn't the reason. 🧠 The real reason was memory. I'd been running with a huge context window, so the AI re-read a giant conversation on every single message. When I stopped doing that, the bill shrank 📉 Same kind of work, a third of the tokens. ✨ The new model helped too, just a lot less than I thought. 💡 Lesson for me: what you feed the AI costs more than which AI you pick. Start fresh often, and keep the conversation lean 🧹 Anyone else measured theirs, or are we all running on vibes? 👇 #claude #claudecode #anthropic #ai #agenticdevelopment #tokens #costs

♻️ 672,007 tonnes of CO2e avoided in a year, just from people buying and selling used…

Tue, 29 Sep 2026 10:35:42 GMT

♻️ 672,007 tonnes of CO2e avoided in a year, just from people buying and selling used stuff across Vend 's marketplaces. FINN.no – mulighetenes marked Blocket Tori.fi dba.dk That's from our 2025 Second Hand Effect Report, released today (PDF attached, longer report link in comments). It's an estimate, not a measurement, and the report says so upfront. I like that. The direction is what matters: second-hand should be the first choice, not the fallback. At home it already is. Most of what we buy comes from FINN, Fretex or the local bruktbutikk. And anything we no longer use that still has life left goes back on FINN. Quite often it's given away for free, but it doesn't go in the bin. The number that surprised me more: 61% of sellers use what they earn to cover everyday household expenses. Selling your old pram isn't only about the climate, for many people it's what keeps the monthly budget working. I get to build the messaging that buyers and sellers use to strike these deals. Nice to see what all those conversations add up to. 💬 What was the last thing you bought second-hand, or gave away instead of throwing out? #SecondHandEffect #CircularEconomy #Recommerce #Sustainability #FINN

When the CEO of the world's largest sovereign wealth fund recommends a Bollywood revenge…

Sun, 27 Sep 2026 04:01:11 GMT

When the CEO of the world's largest sovereign wealth fund recommends a Bollywood revenge saga, a fellow Oslo resident from India has to chime in. 🎬 Nicolai Tangen is right - #Dhurandhar is a spectacle. #RanveerSingh is magnetic, #AdityaDhar holds tension like few in Hindi cinema right now, and across 3.5+ hours it mostly earns its runtime. Special mention for #GauravGera, hats off dude. The "A" certificate is honest, though. Violence isn't just present — it's the film's default setting, dialled so far past impact it tips into numbness. Visceral becomes gratuitous more than once. Where it separates itself is sound. 🎧 Instead of chasing new chartbusters, they reached for songs already etched in collective memory — "Aari Aari," "Tamma Tamma Loge" — and placed them with real care to trigger instant recognition and lift the emotion. A familiar hook lands feeling a new one still has to earn. In a film built on excess, that instinct is what stayed with me. Bloody, loud, a touch too long — and still worth the ticket. Number 2 = 1, as Nicolai put it. Agreement across Norway and India on this one. 💬 If you've seen it — did the old-song callbacks land for you, or did the sheer violence drown them out? #Bollywood #Dhurandhar #FilmReview #SoundDesign #Cinema

🔄 14 months ago, I posted about retiring my sticker collection for a fresh laptop. Clean…

Thu, 24 Sep 2026 00:00:00 GMT

🔄 14 months ago, I posted about retiring my sticker collection for a fresh laptop. Clean slate. Fresh start. All of it. The laptop is fully covered again. A few earned their place — "have fun & get sh**t done" (our department motto), a winter-edition conference tee that beats anything on the market. Lesson from 2025, one year in: peeling stickers off was the ritual, not the point. Laptops get covered again because that's what happens when you actually use them. 👇 If you peeled yours off tomorrow — which sticker do you replace first?

🤖 A year ago I published Taming the Machine — arguing AI regulation should borrow Wall…

Tue, 22 Sep 2026 00:00:00 GMT

🤖 A year ago I published Taming the Machine — arguing AI regulation should borrow Wall Street's playbook 💰 and GDPR's individual safeguards 🔒. Reading it back today, some of it holds up. Some of it clearly doesn't. ✅ What still stands: circuit breakers on high-stakes AI systems, silos to limit blast radius when things fail, mandatory explainability audits, purpose limitation on training data. Regulating for risk containment, not trying to forecast where AI is going. The industry has drifted further toward the failure modes those are designed to contain, not less. ❌ What doesn't: I spent too much of the piece on institutional form — "AI SEC", "AI SEBI" — as if the shape of the regulator was the interesting question. It isn't. Compute governance is the lever that actually moves.

⚡ Two tools that showed up during the last stretch of learning genAI + living inside…

Sat, 19 Sep 2026 00:00:00 GMT

⚡ Two tools that showed up during the last stretch of learning genAI + living inside Claude Code. Both earned their place. For very different reasons. 🎭 Playwright — deterministic, boring, does exactly what you tell it. Not AI. But paired with Claude Code the loop gets unreasonable: agent writes code, Playwright drives the browser, agent watches the trace, fixes, re-runs. No human in the middle. Bonus: per-test video recordings turn "make a demo clip" from an afternoon into a command. 🕷 BrowserMCP — the AI-adjacent one, and where the Skeptic hat goes on. Handing an LLM a live handle on my real browser session — cookies, tabs, everything I'm logged into — is a bad default. Wide privilege surface for a thing whose failure mode is confident wrongness. Nothing I'd point at production. But for scoped, local, "look at this page and click that" work on a throwaway sandbox, it removes the last bit of glue that still needed a human. The rule I'm settling on: Playwright when I need certainty. BrowserMCP when I need speed on work I don't mind blowing up. Not competing — different jobs.

My AI Skeptic & Enthusiast reflex hesitates before installing third-party tools — each…

Thu, 17 Sep 2026 00:00:00 GMT

My AI Skeptic & Enthusiast reflex hesitates before installing third-party tools — each install is another dependency, another attack surface, another thing to keep updated — so I took the shapes people suggested and built local versions instead of adopting the tools. 🎨 Terminal background per codebase. `tbg ` writes an OSC 11 escape to iTerm2, and a hard-coded mapping means the moment I switch codebases the terminal changes colour. Peripheral awareness, zero cognitive load. Rendering the colours is a one-liner — deciding the mapping took longer than writing the script. 🔗 Per-project memory that auto-loads. Every project gets a MEMORY.md that Claude reads at session start. Small typed memory files sit alongside it — user preferences, project state, references — with [[wiki-style]] links between them. Cross-session context is now a filesystem, not a stitched-together conversation history. 🧠 Session sprawl is still only half-solved.

⚡ Most developers, architects and tech folks I know are now running some flavour of CLI…

Tue, 08 Sep 2026 05:02:04 GMT

⚡ Most developers, architects and tech folks I know are now running some flavour of CLI AI orchestrator - Claude Code, OpenCode, Copilot CLI. I started using Claude Code in May after returning from paternity leave, and went all-in after summer. Three things are actively slowing me down: 🧠 Session sprawl. Multiple terminals, multiple sessions, multiple tasks in flight. Keeping track of what's running where takes real cognitive load. 🔗 No shared memory across sessions. When two parallel sessions produce related output, stitching them together is manual work. 🎨 No visual cue for what I'm working in. Production code, a personal prototype, a hackathon repo — they all look the same in the terminal. Easy to type the wrong thing in the wrong window. 👇 If you're further along the Claude Code path than I am, how are you solving these? Less interested in tool names, more in the workflow shape that actually holds up for you. Trying to come up with my own solutions but haven't landed on anything concrete yet -curious to see how much of it overlaps with what the crowd here is already doing. #claude #claudecode #AI #development #cli #terminal

The dress code at JavaZone has two layers: the 👕 tells you where you work, the socks…

Thu, 03 Sep 2026 06:49:30 GMT

The dress code at JavaZone has two layers: the 👕 tells you where you work, the socks tell you what you actually do. 🧦 Day 1 at Vend booth at Javazone was a good one. About 200 people played the Vend quiz and got to know the platform behind FINN.no – mulighetenes marked , Blocket , DBA - Den Blå Avis and Tori.fi, Vend Marketplaces Oy . The all-you-can-eat-and-carry candy bar was the talk of the town 🍬🍭, and the prize we're giving away - easily the most unusual on the floor. Most common questions we heard yesterday: → Wait, is Vend a payments platform? — Close, but no... → Where does Vend come from? — 1999, a Norwegian... → What's the tech stack? — Obviously we use... → Culture, are you hiring? — Both, but the... → What, you have your own podcast? — Ohh ya, we talk about.... → How forward-leaning on AI? — Live, learning, and... Today is the last day. If you're looking for answers about Vend — who we are, what we stand for, how we actually do things — come by booth 51. Same candy bar, same or maybe a different prize, same (well, definitely a few different) questions welcome. Clayton Don Corda #JavaZone #Vend #engineering #AI #javazone2026

1000 deploys a day. Near-zero downtime. Four Nordic marketplaces — finn.no, Blocket, DBA,…

Tue, 01 Sep 2026 14:50:26 GMT

1000 deploys a day. Near-zero downtime. Four Nordic marketplaces — finn.no, Blocket, DBA, Tori — running on one platform. 🚀 Tomorrow at #JavaZone2026, I'll be at the Vend booth wearing my engineer hat — ask me how, and more interestingly, where it hurts. 🛠️ A few things I'm happy to talk shop about: ⚡ How we keep shipping fast without breaking the marketplaces underneath 🤖 How AI is showing up in our day-to-day — what we're trying, what we're learning 🧩 What it's like building for four brands with different users, one codebase Come by booth 51. I'd rather have one honest conversation than ten polite ones. 🤝 #JavaZone2026 #Vend #Engineering #AI #finn

Just shipped a major update to my 2026 FIFA World Cup site: The Historical World Cup Time…

Thu, 16 Jul 2026 10:59:07 GMT

Just shipped a major update to my 2026 FIFA World Cup site: The Historical World Cup Time Machine! 🕒🏆 You can now travel back in time and replay every single knockout tournament from 1954 to 2022 on the exact same interactive, circular bracket. 👉 Travel back in time here: https://lnkd.in/euDj5X78 Pick a legendary year, hover over any nation to light up their path to glory, or hit play to watch the flags physically slide inward as history unfolds match-by-match. When the final match drops, you get the full champion experience: confetti and a stadium roar. 🎉🏟️ What went into this build: - 🗺️ 16 Historical Tournaments: From the 1950s classics to Qatar 2022. - 📐 Adaptive Layouts: Whether it was a 4, 8, or 16-team knockout, the circular UI dynamically adapts while keeping the visual style consistent. - ⏱️ Full Timeline Control: Features a custom scrubber, adjustable playback speeds, and interactive path tracking. - 📱 Mobile-First Design: Fully optimized to run smoothly on any device. #WorldCup #UIUX #DataVisualization #ProductDesign #BuildInPublic

Data visualization challenge: When it comes to sports tournaments, does tradition win, or…

Thu, 09 Jul 2026 07:21:57 GMT

Data visualization challenge: When it comes to sports tournaments, does tradition win, or is it time for a redesign? 🧐⚽ I’ve been building a fan companion app for the 2026 FIFA World Cup, and I’ve implemented two completely different ways to view the knockout stages. Now, I want your unfiltered product and UI/UX feedback. Which layout do you actually prefer using? 1️⃣ The Classic Horizontal Tree Bracket - The Vibe: Traditional, left-to-right, familiar. - The Reality: It's a gold standard for a reason, but it requires a lot of horizontal scrolling, and text can get cramped on smaller mobile viewports. 🔗 Check it out here: https://lnkd.in/eAqVm7mv 2️⃣ The Interactive Radial (Circular) Bracket - The Vibe: Concentric rings, immersive, modern. - The Reality: All 32 teams start on the outer edge and advance inward toward the center trophy. Eliminated teams instantly fade to gray, making it incredibly easy to see the surviving tournament paths at a single glance. 🔗 Check it out here: https://lnkd.in/euDj5X78 Take a quick look at both variants on your phone or desktop, let me know WHY you chose your option in the comments! 👇 #UIUX #ProductDesign #DataVisualization #WorldCup2026 #WebDevelopment #BuildInPublic

You asked, I delivered! 🚀 A brand new feature just dropped on my 2026 FIFA World Cup…

Wed, 08 Jul 2026 14:56:19 GMT

You asked, I delivered! 🚀 A brand new feature just dropped on my 2026 FIFA World Cup site: Replay Mode! 🕒⚽ 👉 Try it here: https://lnkd.in/euDj5X78 Think of it as an interactive, match-by-match time-lapse for the entire tournament: Watch history unfold: Hit play and watch teams physically slide from the outer ring toward the center trophy as they win chronologically 🏆 Pop-up drama: A clean matchcard surfaces live scores and instant official FIFA highlights for every single game 🎥 Total control: Scrub through the timeline like a video track, jump between stages (R32, R16, QF), and adjust the playback speed ⚡ Check out the video below to see it in action. What do you think of this layout? 👇 #WorldCup2026 #UIUX #DataVisualization #BuildInPublic #ProductDesign

Been building a companion site for the 2026 FIFA World Cup — and two updates today…

Wed, 08 Jul 2026 06:31:12 GMT

Been building a companion site for the 2026 FIFA World Cup — and two updates today finally made it click as the one-stop tournament view I've been chasing. ⚽ Check it out: https://lnkd.in/euDj5X78 Instead of a traditional tree, the knockout stage is designed as a circular bracket. 32 teams line the outer ring, the trophy sits in the center, and every nation stands the exact same distance from glory. Here is what went live today: 🎨 Visual Elimination Paths: -Only teams still alive in the tournament stay in full color. Eliminated nations—and a team’s own past tournament positions—instantly fade to gray. At a glance, the remaining path to the final beautifully isolates itself. 🎥 Deterministic Match Highlights: Every single played match now links directly to its official FIFA highlights. - Primary: Canonical match highlights via the FIFA YouTube channel (strict filters: no Alt Casts, no short reels). - Secondary Fallback: If YouTube hasn't uploaded yet, it automatically falls back to FIFA's official match report article. - Right now, 79 of 92 played matches are covered, with a clientless data refresh happening every hour. Clicking any flag slides open a comprehensive Journey Modal—showing a nation's complete group + knockout run, with video highlights baked into every single row. This is a 100% fan-made, open-source passion project. If you want to poke at the code, break things, or contribute, issues and PRs are incredibly welcome! https://lnkd.in/eVGCNXdt Drop your thoughts or bracket predictions below! 👇 #WorldCup2026 #ReactJS #ClaudeCode #OpenSource #WebDevelopment #SideProject

Yesterday, I saw Charlie Lambropoulos sharing Paul Noble’s brilliant circular FIFA World…

Mon, 06 Jul 2026 21:04:48 GMT

Yesterday, I saw Charlie Lambropoulos sharing Paul Noble’s brilliant circular FIFA World Cup bracket concept. The traditional grid has always felt rigid. Paul's circular design completely changed how I envisioned the tournament flow—it felt like a shield wall, unified and strategic. 🛡️ Turning a complex concept like this into a fully responsive app used to take weeks. But we live in the era of augmented engineering. 💻 Leveraging AI, I rapidly forged this idea into an interactive reality. AI handled the heavy lifting of the code, acting as the ultimate multiplier so I could focus on craftsmanship. 🔨 The result is live today: 🌐 https://lnkd.in/euDj5X78 Forged in the Digital Smithy: - Interactive Circular Brackets: Powered by live data from free APIs. - Match Highlights: Sourced directly from official FIFA channels. 🎬 - Upcoming Events: A streamlined schedule for future battles on the pitch. 📅 - 100% Responsive: Built to adapt seamlessly from desktop to mobile screens. 📱 Open to the Realm 🌐 The blueprint is public. I’ve opened the source code for all to study and improve. In the true spirit of collaboration, issues and feedback are most welcome—let's refine this blade together. 🤝 📂 Contribute here: https://lnkd.in/exQuiDnV Test the live build here: https://lnkd.in/euDj5X78 #WebDevelopment #AI #AugmentedEngineering #UIUX #FIFA #BuildInPublic #OpenSource

🧠💥 AI won’t wait—and neither should we.

Thu, 07 Aug 2025 04:48:53 GMT

🧠💥 AI won’t wait—and neither should we. When one software glitch can ground planes and shut down hospitals across 100+ countries (like the recent CrowdStrike outage), imagine what an unchecked black-box AI could do. We’re standing at a crossroads: ➡️ One path leads to innovation with accountability. ➡️ The other? Silent collapse, invisible causes, no one to blame. “Taming the Machine: What AI Regulation Can Learn from Wall Street and GDPR” - In this piece, I explore how we can stop guessing what AI might do—and start building guardrails that actually work. 🔍 Key takeaways: ✅ Agnostic regulation—like the SEC handles markets ✅ Explainability audits—like GDPR ensures individual rights ✅ Circuit breakers, kill switches, bias testing ✅ Why siloed systems matter—and global cooperation matters more 👉 If AI is already making decisions that shape justice, medicine, and markets, don’t we owe it the same scrutiny we give stock trades? Let’s build resilience, not just regulation. Feedback and discussion welcome 👇 #AI #ArtificialIntelligence #AIEthics #TechPolicy #GDPR #WallStreet #AIRegulation #MachineLearning #Leadership #DigitalSovereignty #CrowdStrike #DataGovernance