Charles Wilson

@santifer Proud to be part of @santifer 's #CareerOps project. Power to the candidates! I hope an EN-translated version is released soon. #JobSearchTools #JobSearch

WIRED 🇬🇷 dropped a deep dive on career-ops today. “Companies use AI to filter candidates. I just gave candidates AI to choose companies” 42K+ stars. Free, open source. wired.com.gr/article/to-ai-…

@championswimmer When AI model providers raise their fees, will more companies shift to on-prem open models?

We were being subsidised this much? Wow. Cursor claiming the $200 plan is actually $5000 was actually true it seems.

@benln The real hard part becomes making the company legible enough for AI to actually help. Most companies are full of undocumented decisions, hidden context, and, most importantly, knowledge stuck in people's heads.

@ttorres The hidden product decision is whether the system is allowed to invent structure or only preserve intent. Once capture starts smuggling in planning, people stop speaking naturally and start prompt-engineering the inbox.

@aakashgupta "Security/infra will be [critical]... every vibe-coded feature still needs someone awake at 3am when it hallucinates...More output means more surface area to secure, more [infra] to scale, more edge cases to catch. The ratio of builders to maintainers is inverting." @aakashgupta

This framework is already playing out in real headcount data. The people who think "vibe coder/PM/slop cannon" makes them safe are the most exposed. Shopify went from 11,600 employees to 8,100 in two years. Lütke now requires teams to prove AI can't do the job before requesting new headcount. The vibe coders are the ones doing the proving. Klarna replaced 700 customer service agents with an AI chatbot and publicly said they're heading toward 2,000 total employees. The roles that survived: security and infra (keeping it running), the hot people (humans customers still want to talk to), and the grown ups (lawyers making sure the chatbot doesn't promise something legally actionable). Security/infra will be the highest paid of the four. Every vibe-coded feature, every agent workflow, every AI-generated module still needs someone awake at 3am when it hallucinates in production. More output means more surface area to secure, more infrastructure to scale, more edge cases to catch. The ratio of builders to maintainers is inverting. "Hot people" is the one people laugh at, but Apple figured this out 20 years ago. The entire Apple Store exists to make you feel something. The margin lives in the feeling. As AI handles more transactional work, the premium on human warmth goes up. The missing 5th job: the person who decides which AI to buy. Enterprise AI tooling is on track to pass $300B annually. Someone has to evaluate whether Claude or GPT or Gemini works for their stack. That role barely exists in org charts yet.

Claude Code doesn't show you how many tokens you're using for subscriptions. No breakdown by model. No breakdown by project. Just a progress bar that says "63% used." So I built a local dashboard that reads the files Claude Code already writes to your machine. Turns out every session, every turn, every token is logged to ~/.claude/projects/ in JSONL files. Input tokens, output tokens, cache reads, cache creation, model name, timestamp. It's all there. You just can't see it. My numbers over the last 30 days: 440 sessions. 18,000 turns. $1,588 in API-equivalent costs. On one day, the cache spiked to 700M tokens - visible cache bug, two days in a row. The dashboard scans those local files, builds a SQLite database, and serves charts on localhost:8080. Filter by model (Opus, Sonnet, Haiku). Filter by time range (7d, 30d, 90d, all time). Cost estimates based on current Anthropic API pricing. Works retroactively. First run processes your entire Claude Code history. Install: git clone github.com/phuryn/claude-… cd claude-usage python3 cli.py dashboard Windows: use python instead of python3. Zero dependencies. Python standard library only. Open source, MIT. Star it. Fork it. Make it your own.

Great article from @ttorres - "You can't win opinion battles. But you can bring information your stakeholders don't have—insights from customer interviews, data from assumption tests, patterns in the opportunity space that they haven't seen." buff.ly/tbtQANk

Ever launch a new product feature that seemed solid—until a small, overlooked detail broke everything? In this episode, Teresa Torres and Petra Wille share their real-world experiences wrangling global invoicing and taxes while running small, international businesses. What starts as a rant about EU tax compliance turns into a sharp product lesson: how failing to map the entire path to customer value—down to the tiniest regulatory requirement—can kill your product’s usefulness. Whether you’re shipping code or selling courses, this conversation will remind you that sweating the details isn’t about perfectionism—it’s about ensuring your product actually delivers value at the moment that matters. In this episode: - The nightmare of global invoicing for small online businesses - Why even big platforms (like Squarespace and Teachable) miss the mark on EU tax compliance - How Petra and Teresa navigated invoicing across borders with Ableify and LearnWorlds - The key difference between meeting regulations and meeting customer needs - What product teams can learn from regulatory edge cases - How missing a single detail can block the “moment of value creation” - Why story mapping is critical for finding gaps between “we shipped it” and “customers got value” Key takeaways: - Customers define value, not your compliance checklist. - Regulatory work still requires discovery—you can’t skip understanding user needs. - The path to value doesn’t end when your feature works; it ends when your customer succeeds. - “Sweating the details” isn’t micromanagement—it’s good product management. Memorable quotes: “If you don’t sweat the details, people choose other platforms.” — Petra Wille “It’s not a little detail when your client won’t pay the invoice.” — Teresa Torres Watch or listen to this episode: YouTube: buff.ly/DRVEfiJ Apple Podcasts: buff.ly/x24rdd7 Spotify: buff.ly/7nkwwSJ

"Discovery doesn't come at the expense of delivery. Both activities should happen in tandem to achieve the best results." 🔄 Explore the 7 most common reasons product teams struggle to make time for discovery: 1. Too focused on delivery activities 📊 2. Excessive stakeholder management 👥 3. Constant cross-team coordination 🤝 4. Overwhelmed by product support tasks 🆘 5. Trapped in endless meetings ⏰ 6. Discovery not valued by the organization 🏢 7. Discovery process itself takes too long 🐌 Learn practical tips to overcome these challenges and integrate discovery into your workflow. Watch the video or read the transcript: buff.ly/CpcndG2 🤔 Which of these reasons resonates most with your team's experience? Share your thoughts in the comments below.

@DanBrownUSA @labsdotgoogle That is a pretty engaging conversation (between two AI personas). Google's #NotebookLM would be really helpful for people who want to learn about topics that haven't been summarized yet.

@labsdotgoogle I put in my dad's PHD Thesis and it created this podcast. Im speechless...

Imagine listening to a lively discussion about your latest research project, summarized and analyzed by two AI hosts 🎧That's the power of #NotebookLM's Audio Overview. Upload PDFs, Google Docs, Slides, and try it for yourself. ✨Learn more at: blog.google/technology/ai/…

CapitalOne doesn't offer a chart that shows where I spend my money. Seems weird, because all the purchased are categorized in the transaction list. C'mon @CapitalOne ! You can do better.

@heyandras @mijustin Wow, the breadth of comparability looks huge- many languages/ hosting environments. What would be the main reason people want this? Controlling costs? (I’ve heard scaling Heroku can get pricey.)

@mijustin built for 3 years as a side-project and 8 months full-time: coolify.io 🙂

@morajabi @mijustin @NoorChat Wow, this looks pretty full-featured, what problem drove you to make it?

@mijustin @NoorChat

@mijustin Podgagement® helps podcasters who feel stuck to engage their audiences and grow their podcasts. podgagement.com

"Make a multiplayer drawing app where the strokes appear on everyone else's screens in realtime. let user pick a name and color. save users to db on login" 2m48s, no bugs: - users & drawings persist to sqlite - socket multiplayer one-shot video (claude 3 opus) demo at end

@jkjenkinney @msabcleek Thank you for producing "Monaea, a 2020 Diary"! I shared a portion of the piece to inspire students about youth podcasting at Bullitt School District's Riverview Opportunity Center (a virtual high school). You rock! bullittschools.org/o/roc

"go.ted.com/6WzM via @TEDTalks "Are you a soldier, prone to defending your viewpoint at all costs -- or a scout, spurred by curiosity? @juliagalef

Calling all podcast-app developers! Please give me your best screenshots of Podcasting 2.0 features displaying in your apps! (No editing, and device framing is optional.)

NEW: an ideological divide is emerging between young men and women in many countries around the world. I think this one of the most important social trends unfolding today, and provides the answer to several puzzles.

Postgres (serverless) is too good. SSR that feels as fast as local data. next-ai-news.vercel.app/search