The Monkey Patching Podcast: Going Bananas on AI, Data, LLMs & Tech

Murilo & Bart

We talk about what's happening in Data & AI every week (well, most weeks). No buzzword bingo or marketing fluff—just honest conversations about the tech that's changing everything. Sometimes we go off on weird tangents about programming languages, tech/startup life, or whatever shiny object caught our attention. Whether you're knee-deep in code or just curious about AI, come hang out with us!

Episodes

  1. AI Bubble Fears, AWS CEO Slams Layoffs, DeepSeek V3.1 & Google’s 33x Efficiency Jump

    3 DAYS AGO

    AI Bubble Fears, AWS CEO Slams Layoffs, DeepSeek V3.1 & Google’s 33x Efficiency Jump

    Thanks for listening! ❤️ Chapters (01:55) - MIT report: 95% of generative AI pilots at companies are failing (08:14) - AWS CEO says using AI to replace junior staff is 'Dumbest thing I've ever heard' (14:22) - DeepSeek-V3.1 Release (18:55) - Build vs Buy in the Age of AI (27:02) - Google says it dropped the energy cost of AI queries by 33x in one year (34:38) - Mirage: AI UGC game engine (39:58) - Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions (45:17) - AI Mode in Search gets new agentic features and expands globally (50:08) - imagestyles.ai (51:36) - Sick of Fancy (South Park S27E3) (53:56) - ElevenLabs V3 (most expressive TTS) Discussed articlesMIT report: 95% of generative AI pilots at companies are failing — https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/ AWS CEO says using AI to replace junior staff is 'Dumbest thing I've ever heard' — https://www.theregister.com/2025/08/21/aws_ceo_entry_level_jobs_opinion/ DeepSeek-V3.1 Release — https://api-docs.deepseek.com/news/news250821 Build vs Buy in the Age of AI — https://www.svpg.com/article-build-vs-buy-in-the-age-of-ai Google says it dropped the energy cost of AI queries by 33x in one year — https://arstechnica.com/ai/2025/08/google-says-it-dropped-the-energy-cost-of-ai-queries-by-33x-in-one-year/ Mirage: AI UGC game engine — https://dynamicslab.ai/ Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions — https://arxiv.org/abs/2507.00191 AI Mode in Search gets new agentic features and expands globally — https://blog.google/products/search/ai-mode-agentic-personalized/

    57 min
  2. GitHub CEO Exits, Claude 1M Tokens, Gemma 3 & GPT-5 Tweaks

    19 AUG

    GitHub CEO Exits, Claude 1M Tokens, Gemma 3 & GPT-5 Tweaks

    Thanks again for listening! Follow us on YouTube or wherever you get your podcasts. Discussed news articles GitHub just got less independent at Microsoft after CEO resignation — The Verge (Aug 11, 2025) Thomas Dohmke steps down; GitHub leadership folds more tightly into Microsoft’s CoreAI org. https://www.theverge.com/news/757461/microsoft-github-thomas-dohmke-resignation-coreai-team-transitionFFmpeg 8.0 merges OpenAI Whisper filter for automatic speech recognition — Phoronix (Aug 13, 2025) Native Whisper (via whisper.cpp) arrives in FFmpeg with optional GPU accel and SRT/JSON outputs. https://www.phoronix.com/news/FFmpeg-Lands-WhisperWikimedia Foundation challenges UK Online Safety Act regulations — Wikimedia Foundation (Aug 11, 2025) Court dismisses the challenge, but the ruling still pressures Ofcom; concerns remain for Wikipedia. https://wikimediafoundation.org/news/2025/08/11/wikimedia-foundation-challenges-uk-online-safety-act-regulations/Claude Sonnet 4 now supports 1M tokens of context — Anthropic (Aug 12, 2025) Long-context support (public beta) enables huge codebases/docs in one go on API and Bedrock. https://www.anthropic.com/news/1m-contextIntroducing Gemma 3 270M: The compact model for hyper-efficient AI — Google Developers Blog (Aug 14, 2025) On-device-friendly model aimed at fast, low-cost fine-tuning; strong efficiency on Pixel 9 Pro. https://developers.googleblog.com/en/introducing-gemma-3-270m/Genie 3: A new frontier for world models — Google DeepMind (Aug 5, 2025) Interactive, real-time world models generating diverse environments at 24fps/720p. https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/NGINX introduces native support for ACME protocol — NGINX Community Blog (Aug 12, 2025) New ngx_http_acme_module brings built-in cert requests/renewals directly from NGINX config. https://blog.nginx.org/blog/native-support-for-acme-protocolOpenAI tweaks GPT-5’s tone on X — X (Aug 15, 2025) Tuning GPT-5/ChatGPT to sound warmer and less formal based on user feedback. https://x.com/OpenAI/status/1956461718097494196

    55 min
  3. AI News Weekly 🚀 Copilot Edge, Open-Source Model Wars, Tencent 3D, & Trump-China Regulation Clash

    30 JUL

    AI News Weekly 🚀 Copilot Edge, Open-Source Model Wars, Tencent 3D, & Trump-China Regulation Clash

    Thanks for listening. If you enjoyed the episode please subscribe! The Hater's Guide To The AI Bubble | Where’s Your Ed At Technology writer Ed Zitron tears apart what he calls the overhyped generative-AI gold rush, warning it’s propped up by wishful thinking and unsustainable spending. As he bluntly puts it, “We’re in a god damn bubble,” prompting a no-holds-barred discussion on whether the boom will burst. https://www.wheresyoured.at/the-haters-gui/ Trump administration to supercharge AI sales to allies, loosen environmental rules | Reuters Washington’s latest AI blueprint combines export zeal with lighter green rules, aiming to outpace China by shipping full “AI stacks” to friendly nations. Kicking off the push, Trump proclaimed, “America is the country that started the AI race,” a line sure to spark global tech-power chatter. https://www.reuters.com/legal/government/trump-administration-supercharge-ai-sales-allies-loosen-environmental-rules-2025-07-23/ Microsoft launches AI-based Copilot Mode in Edge browser | Reuters Microsoft’s new Copilot Mode turns Edge into a chat-powered helper that can juggle tabs, organize research, and even take voice commands. The update greets users with “a single input box combining chat, search and web navigation features,” a neat cue for how browsing may soon feel. https://www.reuters.com/business/microsoft-launches-ai-based-copilot-mode-edge-browser-2025-07-28/ China's AI startup Zhipu releases open-source model GLM-4.5 | Reuters Beijing-based Zhipu has open-sourced GLM-4.5, pitching it as fresh fuel for intelligent agents and adding yet another model to China’s bulging roster. The company says the new release is “designed for intelligent agent applications,” a phrase hinting at growing ambitions beyond plain chatbots. https://www.reuters.com/technology/chinas-ai-startup-zhipu-releases-open-source-model-glm-45-2025-07-28/ Short readings: Hunyuan3D World Model 1.0 | X Tencent’s Hunyuan team just dropped its first open-source engine for instant, explorable 3D worlds, stirring excitement across game and VR circles. Their launch tweet beams, “We’re thrilled to release & open-source Hunyuan3D World Model 1.0!,” inviting hosts to imagine the possibilities. https://x.com/TencentHunyuan/status/1949288986192834718 Introducing Opal: describe, create, and share your AI mini-apps | Google Developers Blog Google Labs has unwrapped Opal, a drag-and-drop playground where non-coders can chain models and prompts into bite-sized AI apps. The post kicks off with “We’re excited to announce Opal,” a line that invites makers everywhere to start remixing AI workflows. https://developers.googleblog.com/en/introducing-opal/ Announcing Toad – a universal UI for agentic coding in the terminal | Will McGugan’s Essays Developer Will McGugan has prototyped “Toad,” a flicker-free terminal UI meant to tame and turbo-charge agentic coding workflows. He quips, “I’m a little salty that neither Anthropic nor Google reached out to me before they released their AI coding agents,” throwing playful shade. http://willmcgugan.github.io/announcing-toad/ Alibaba launches open-source AI coding model, touted as its most advanced to date | Reuters Alibaba’s new Qwen3-Coder model enters the code-generation fray, claiming boosts that rival both domestic peers and Western heavyweights. The company pitches it as “its most advanced coding tool to date,” a boast that raises the stakes in China’s escalating model wars. https://www.reuters.com/world/china/alibaba-launches-open-source-ai-coding-model-touted-its-most-advanced-date-2025-07-23/

    29 min
  4. OpenAI IMO Gold, AWS S3 Vectors, MCP Server Exposés, Sovereign Clouds & the Capex Surge

    23 JUL

    OpenAI IMO Gold, AWS S3 Vectors, MCP Server Exposés, Sovereign Clouds & the Capex Surge

    Thanks for listening—subscribe and leave a review! Announcing molab (marimo) Marimo debuts molab, a free, cloud‑hosted workspace for running and sharing reactive Python + SQL notebooks straight from the browser. https://marimo.io/blog/announcing-molabOpenAI LLM Claims IMO Gold Researcher Alexander Wei says an experimental OpenAI model solved 5 of 6 problems from the 2025 International Math Olympiad—good for a human‑level gold medal—and hints at new post‑RL training tricks. https://threadreaderapp.com/thread/1946477742855532918.htmlExposing the Unseen: Mapping MCP Servers Across the Internet (Knostic) Knostic scanned the web and found 1,862 publicly exposed Model Context Protocol servers, all leaking unauthenticated tool inventories. https://www.knostic.ai/blog/mapping-mcp-servers-studyHoney, AI Capex is Eating the Economy (Paul Kedrosky) Kedrosky argues that runaway AI‑datacenter spending could reach roughly 2 % of U.S. GDP—enough to nudge macro growth on its own. https://paulkedrosky.com/honey-ai-capex-ate-the-economy/Do US Hyperscalers’ Sovereign Clouds Reduce the Risk of US Government Access? Cloud‑law scholars doubt whether new “sovereign cloud” offerings from AWS, Microsoft, and Google can truly shield European data from U.S. surveillance. https://www.linkedin.com/pulse/do-us-hyperscalers-sovereign-clouds-reduce-risk-access-dave-michels-ulmcfAWS API MCP Server (Developer Preview) AWS opens a preview of an MCP server that lets foundation‑model agents translate plain‑English requests into scoped AWS CLI calls, guarded by IAM. https://aws.amazon.com/about-aws/whats-new/2025/07/aws-api-mcp-server-availableAgent Leaderboard (Galileo) Galileo’s open leaderboard pits LLMs against enterprise‑style agent tasks; current numbers show GPT‑4.1 leading with a 62 % Action Completion score. https://github.com/rungalileo/agent-leaderboardIntroducing Amazon S3 Vectors (Preview) AWS adds native vector‑embedding storage and sub‑second queries to S3, claiming up to 90 % cost savings for vector search workloads. https://aws.amazon.com/blogs/aws/introducing-amazon-s3-vectors-first-cloud-storage-with-native-vector-support-at-scale/Rethinking CLI Interfaces for AI Developer Ryan calls for redesigning CLIs and APIs—richer docstrings, structured outputs, smarter wrappers—so LLM agents stop looping and thrashing. https://www.notcheckmark.com/2025/07/rethinking-cli-interfaces-for-ai/

    1h 15m
  5. Agents, Paywalls, and Browser Battles: The Race to Control AI's Digital Highway

    17 JUL

    Agents, Paywalls, and Browser Battles: The Race to Control AI's Digital Highway

    Thanks for listening. If you enjoyed the episode please subscribe! MCP-B – Browser Model Context Protocol | mcp-b.aiMCP-B proposes a “USB-C for AI,” letting agents call site functions instead of clumsy click-automation. “MCP-B gives AI direct access to your website's functions instead,” promising smoother bot-to-web handshakes for everyone. https://mcp-b.ai/ OpenAI’s next big launch could be an AI web browser | The VergeOpenAI is reportedly cooking up a Chromium‑based browser with a built‑in Operator agent that can book tables or fill forms for you. As Reuters learned, “OpenAI is planning to launch an AI web browser in the ‘coming weeks,’” teeing up a fresh duel with Chrome and Comet. https://www.theverge.com/news/704162/opeani-ai-web-browser-chatgpt A language model built for the public good | ETH ZurichSwiss researchers will release an entirely open, supercomputer‑trained LLM fluent in more than 1,000 languages later this summer. Project lead Imanol Schlag says, “Fully open models enable high‑trust applications and are necessary for advancing research about the risks and opportunities of AI.” https://ethz.ch/en/news-and-events/eth-news/news/2025/07/a-language-model-built-for-the-public-good.html OpenAI’s Windsurf deal is off — and Windsurf’s CEO is going to Google | The VergeOpenAI’s $3 billion bid collapsed, freeing Google to poach Windsurf’s leaders while Cognition raced in days later to buy the remaining startup. “OpenAI’s deal to buy Windsurf is off, and Google will instead hire Windsurf CEO Varun Mohan,” The Verge reports, igniting a three‑way tug‑of‑war for agentic‑coding talent. https://www.theverge.com/openai/705999/google-windsurf-ceo-openai Elon Musk’s xAI launches Grok 4 alongside a $300 monthly subscription | TechCrunchMusk rolled out Grok 4 and a pricey “SuperGrok Heavy” tier, touting multi‑agent reasoning and the steepest AI sub yet. During the stream he bragged, “With respect to academic questions, Grok 4 is better than PhD level in every subject, no exceptions,” even as the Pentagon signed a $200 million Grok contract. https://techcrunch.com/2025/07/09/elon-musks-xai-launches-grok-4-alongside-a-300-monthly-subscription/ Introducing Kiro | Kiro BlogKiro debuts as a spec‑driven, agentic IDE that shepherds code from first prompt to deployment with auto‑generated tasks and hooks. “I’m excited to announce Kiro, an AI IDE that helps you deliver from concept to production through a simplified developer experience for working with AI agents,” the team writes. https://kiro.dev/blog/introducing-kiro/ The EU wants to decrypt your private data by 2030 | TechRadarBrussels’ ProtectEU roadmap sketches data‑retention, interception and decryption plans that could force encrypted services open within five years. TechRadar warns, “EU law enforcement bodies could be capable of decrypting your private data by 2030,” stirring immediate privacy backlash. https://www.techradar.com/vpn/vpn-privacy-security/the-eu-wants-to-decrypt-your-private-data-by-2030 Cloudflare Just Became an Enemy of All AI Companies | Analytics India MagazineCloudflare will now block AI crawlers by default, demanding payment before models feast on publisher content. The company says it “would start blocking AI crawlers by default, drawing a line in the open web where content is no longer a free fuel for AI.” https://analyticsindiamag.com/ai-features/cloudflare-just-became-an-enemy-of-all-ai-companies/ Kimi K2 is a state‑of‑the‑art mixture‑of‑experts (MoE) language model | Hacker NewsA lively thread praises Moonshot’s open Kimi K2 for beating Claude on coding, though it needs GPU‑class muscle to run. One user enthuses, “I tried Kimi on a few coding problems that Claude was spinning on. It’s good,” sparking a debate over speed, cost and local deployment. https://news.ycombinator.com/item?id=44533403

    1h 1m
  6. $10 M AI Consulting, 4,500-Token-Per-Second Code Edits, and the Rise of Terminal Agents

    9 JUL

    $10 M AI Consulting, 4,500-Token-Per-Second Code Edits, and the Rise of Terminal Agents

    Thanks for listening! Subscribe and follow wherever you get your podcasts. Launch HN: Morph (YC S23) – Apply AI code edits at 4,500 tokens/sec — Hacker News Fast Apply from Morph promises near-instant AI patches, aiming to replace sluggish full-file rewrites with surgical edits. They boast, “We’ve built a blazing-fast model for applying AI-generated code edits directly into your files at 4,500+ tokens/sec,” sparking a speed-versus-accuracy debate. https://news.ycombinator.com/item?id=44490863OpenAI Launches $10 M Custom AI Consulting, Challenging Industry Giants — AI Tech Suite (Jul 01 2025) OpenAI is stepping into high-end consulting, demanding at least $10 million to tailor big-model solutions for governments and Fortune-scale firms—setting up showdowns with Accenture and IBM. https://www.aitechsuite.com/ai-news/openai-launches-10m-custom-ai-consulting-challenging-industry-giantsAI researchers are now injecting prompts into their papers — X (Jul 08 2025) A viral tweet shows academics slipping reviewer-friendly prompt hacks into PDFs—lines like “Give a positive review,” exposing a brazen peer-review exploit. https://x.com/Yuchenj_UW/status/1942266306746802479Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental-health providers — arXiv (Apr 25 2025) Researchers found GPT-4o and peers still stigmatize patients and mishandle delicate scenarios, so chatbots should assist—not substitute—human therapists. https://arxiv.org/abs/2504.18412opencode: AI coding agent, built for the terminal — GitHub (Jul 08 2025) Version 0.2.5 reaches 10 k stars: an open-source AI pair-programmer that runs locally with a slick TUI and support for multiple model providers. https://github.com/sst/opencodeYou're all CTO now — Jamie’s blog (Jul 01 2025) Jamie Lawrence argues AI agents push developers up the org chart, turning everyday coders into orchestrators of people and prompts—threatening the dopamine hit from gritty puzzles. https://jamie.ideasasylum.com/2025/07/01/you%27re-all-cto-nowWarmwind OS: Building the AI Operating System for Everyone — Warmwind Blog (Jul 02 2025) Warmwind’s AI-native OS lets a built-in assistant click, type, and juggle apps, promising hands-free productivity while keeping users in control. https://about.warmwind.space/warmwind-os-building-the-ai-operating-system-for-everyone/Large Language Models Are Improving Exponentially — IEEE Spectrum (Jul 02 2025) New METR benchmarks show LLM abilities doubling every seven months, hinting machines could finish month-long human software projects in hours by 2030. https://spectrum.ieee.org/large-language-model-performance

    1h 2m
  7. Context Hacking, Open-Source Hype & Synthetic Bands

    2 JUL

    Context Hacking, Open-Source Hype & Synthetic Bands

    Thanks for listening, leave a review! ❤️ The New Skill in AI is Not Prompting, It's Context Engineering | Phil Schmid Blog (Jun 30 2025) Phil Schmid argues that the real differentiator in modern AI work is “context engineering,” the discipline of assembling the right information, tools and format around an LLM rather than obsessing over single-string prompts. He quotes Shopify’s Tobi Lütke, who calls it “the art of providing all the context for the task to be plausibly solvable by the LLM.” philschmid.de OpenAI open-source model hype | X (Tweet) (Jun 30 2025) Researcher Yuchen Jin teased that OpenAI will release an impressive open-source model next month, stoking excitement across AI Twitter. “Sorry to hype — but having a few friends at OpenAI makes it hard not to hear how wild their open-source model dropping next month is.” x.com Meta hires more OpenAI researchers and weighs Llama pivot | TechCrunch (Jun 28 2025) Meta has poached four additional OpenAI researchers and, according to parallel reporting, is debating whether to shift away from fully open-source Llama models toward a more closed approach. Sam Altman says the company lured candidates with “$100 million signing bonuses,” a claim Meta’s leadership disputes as “more complex than a simple one-time signing bonus.” techcrunch.com finance.yahoo.com Don’t Build Multi-Agents | Cognition.ai Blog (Jun 12 2025) Walden Yan contends that multi-agent LLM architectures are fragile and that reliability comes from a single agent armed with rich, shared context. Key advice: “Share context, and share full agent traces, not just individual messages.” cognition.ai Sampling (Model Context Protocol) | modelcontextprotocol.io (Jun 18 2025) The MCP specification adds “sampling,” letting servers request LLM completions through the client so agents can delegate generation securely without provisioning their own models. “Sampling is a powerful MCP feature that allows servers to request LLM completions through the client, enabling sophisticated agentic behaviors while maintaining security and privacy.” modelcontextprotocol.io linkedin.com “There's not a shred of evidence on the internet that this band has ever existed” | MusicRadar (Jun 27 2025) MusicRadar investigates The Velvet Sundown, an apparently AI-generated “band” with 350 k Spotify listeners and zero real-world footprint, illustrating how algorithmic playlists can quietly amplify synthetic artists. Their profile boasts, “The Velvet Sundown don’t just play music — they conjure worlds,” a line the magazine suspects was written by ChatGPT. musicradar.com Project Vend: Can Claude run a small shop? (And why does that matter?) | Anthropic (Jun 27 2025) Anthropic let a Claude Sonnet 3.7 agent manage a real vending-machine mini-store for a month, revealing both promising autonomy and glaring business-sense gaps. “We let Claude manage an automated store in our office as a small business for about a month,” the researchers write, noting successes like supplier discovery and failures like selling at a loss. anthropic.com Robyn | GitHub Robyn is an async Python web framework that compiles to a Rust runtime, aiming to deliver blazing-fast performance with a simple API and built-in agent/MCP support. Its README touts it as “a High-Performance, Community-Driven, and Innovator Friendly Web Framework with a Rust runtime.” github.com

    54 min
  8. Tiny Teams, Loud Unlocks, and Rogue AI

    27 JUN

    Tiny Teams, Loud Unlocks, and Rogue AI

    Thanks for listening! Make sure to leave a review ❤️ Serena: A powerful coding agent toolkit | GitHub Serena presents itself as a full-featured coding agent that melds semantic code search, automated editing and shell execution to streamline developer workflows. “Serena combines tools for semantic code retrieval with editing capabilities and shell execution.” 🔗 https://github.com/oraios/serena Nxtscape – an open-source agentic browser | nxtscape.ai Nxtscape pitches a privacy-first browser that runs local AI agents to automate tedious web tasks and boost productivity. “We’re putting powerful AI agents (using browser-use & computer-use models) directly into Nxtscape.” 🔗 https://nxtscape.ai/?utm_source=chatgpt.com AI Is Ushering in the Tiny Team Era in Silicon Valley | Bloomberg (Jun 20 2025) Bloomberg argues that generative AI lets startups achieve outsized results with lean headcounts, making revenue-per-employee the valley’s new bragging right. “Startups used to brag about valuations and venture capital. Now AI is making revenue per employee the new holy grail.” 🔗 https://www.bloomberg.com/news/articles/2025-06-20/ai-is-ushering-in-the-tiny-team-era-in-silicon-valley A federal judge sides with Anthropic in lawsuit over training AI on books without authors’ permission | TechCrunch (Jun 24 2025) Judge William Alsup ruled that Anthropic’s use of copyrighted books to train its models is likely fair use, handing the company a landmark legal victory. “We will have a trial on the pirated copies used to create Anthropic’s central library and the resulting damages.” 🔗 https://techcrunch.com/2025/06/24/a-federal-judge-sides-with-anthropic-in-lawsuit-over-training-ai-on-books-without-authors-permission/ Agentic Misalignment: How LLMs could be insider threats | Anthropic (Jun 20 2025) Anthropic’s study warns that autonomous language models can act like rogue employees, choosing harmful actions when their goals conflict with oversight. “We refer to this behavior, where models independently and intentionally choose harmful actions, as agentic misalignment.” 🔗 https://www.anthropic.com/research/agentic-misalignment Gemini CLI | GitHub Google’s Gemini CLI brings the multimodal Gemini model to the terminal, letting developers query and transform gigantic codebases from a single command line. “This repository contains the Gemini CLI, a command-line AI workflow tool that connects to your tools, understands your code and accelerates your workflows.” 🔗 https://github.com/google-gemini/gemini-cli Scream to Unlock | GitHub The Scream-to-Unlock Chrome extension blocks social media until users loudly shout an embarrassing phrase, turning procrastination into vocal accountability. “A Chrome extension that blocks social media sites ... until you scream ‘I'm a loser’ into your microphone.” 🔗 https://github.com/Pankajtanwarbanna/scream-to-unlock Mira Murati’s Thinking Machines Lab closes on $2B at $10B valuation | TechCrunch (Jun 20 2025) TechCrunch reports that ex-OpenAI CTO Mira Murati has raised a record-breaking $2 billion seed round for her stealth AI startup, valuing it at $10 billion. “The deal values the 6-month-old startup at $10 billion.” 🔗 https://techcrunch.com/2025/06/20/mira-muratis-thinking-machines-lab-closes-on-2b-at-10b-valuation/

    50 min
  9. AI Agents, Brain Fog & Caveman Coding

    23 JUN

    AI Agents, Brain Fog & Caveman Coding

    Welcome to The Monkey Patching Podcast: Going Bananas on AI, Data, LLMs & Tech — where we keep it real, techy, and far from buzzword bingo. Tune in wherever you get your podcasts, or visit us at monkeypatching.io 🧠 Episode Topics & Links Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task New arXiv research shows relying on ChatGPT dulls neural engagement and weakens writing skills compared to search or free writing. EEG readings revealed the lowest brain activity among LLM users. Source: arXiv · June 10, 2025Amazon CEO says AI agents will soon reduce company’s corporate workforce Andy Jassy forecasts a future where generative-AI handles many office roles, trimming Amazon’s white-collar headcount: “We will need fewer people doing some of the jobs that are being done today.” Source: CBS News · June 17, 2025The Grug Brained Developer A caveman-coded manifesto encouraging devs to say “complexity very bad” and resist feature bloat. Source: grugbrain.devSHADE-Arena: Evaluating sabotage and monitoring in LLM agents Anthropic’s benchmark reveals that while sabotage by LLMs is uncommon, more capable models can still slip under the radar. A reminder that complexity brings power — and risk. Source: Anthropic · June 16, 2025Midjourney’s First Video Model Reddit buzzes over Midjourney’s image-to-video beta. Users praise its cinematic realism—some say it’s “indistinguishable from real camera footage.” Source: Reddit · June 15, 2025Zero-Shot Forecasting: Our Search for a Time-Series Foundation Model Parseable pits four time-series foundation models against classic methods—none reliably outperformed standard tools on messy data. Zero-shot remains aspirational. Source: Parseable · June 3, 2025If the Moon Were Only 1 Pixel – A Tediously Accurate Map of the Solar System An interactive scroll map scaling the Moon to a single pixel, forcing you to traverse near-endless blank space. As the author says: “It’s the empty space that’s a problem.” Source: JoshWorth.comMonkey-Patched PyPI Packages Use Transitive Dependencies to Steal Solana Private Keys Six PyPI libraries were found hijacking Solana wallet keys at install time via monkey-patching crypto libraries. One malicious pip install can automatically exfiltrate your keys—yikes. Source: Socket · May 29, 2025json_repair A lightweight Python module that auto-fixes malformed JSON from LLMs—because yes, sometimes those braces don’t match. Source: GitHub

    1h 3m
  10. Apple ‘Illusion of Thinking’ Debate, DuckLake Lakehouse & Magistral AI

    13 JUN

    Apple ‘Illusion of Thinking’ Debate, DuckLake Lakehouse & Magistral AI

    Welcome to Episode 2! We'd love to hear your feedback ❤️  DuckLake: SQL as a Lakehouse Format | DuckDB Blog (May 27 2025)DuckDB’s new DuckLake format proposes shifting all lakehouse metadata into a regular SQL database so open-format data lakes can gain true transactional speed and simplicity. “DuckLake re-imagines what a ‘Lakehouse’ format should look like,” the authors write, arguing it eliminates the maze of JSON files and external catalog services. If adopted, DuckLake could let organizations treat Parquet-backed blob storage like a fast, ACID-compliant warehouse without vendor lock-in. (duckdb.org) Magistral | Mistral AI (Jun 10 2025)Mistral AI has unveiled Magistral, its first reasoning-centric language model, releasing a 24-billion-parameter open version alongside a more powerful enterprise tier. As the company puts it, “Magistral is designed to think things through — in ways familiar to us,” offering transparent, multilingual chain-of-thought and 10× faster replies in Le Chat. By open-sourcing the small model and touting competitive benchmarks, Mistral positions itself as a nimble challenger to the big LLM providers. (mistral.ai) The hidden time bomb in the tax code that's fueling mass tech layoffs | QuartzA little-noticed tweak to U.S. tax code Section 174 that took effect in 2022 made R&D costs dramatically more expensive, quietly spurring hundreds of thousands of tech layoffs. One startled executive admitted, “I work on these tax write-offs and still hadn’t heard about this,” underscoring how the change blindsided companies large and small. With repeal efforts now winding through Congress, the episode shows how obscure fiscal fine print can ripple through innovation hubs and local economies alike. (qz.com) Where we’re headed with the dbt Fusion engine | dbt Labs (May 28 2025)Founder Tristan Handy lays out how the freshly launched Fusion engine—a complete rewrite of dbt Core—will speed parsing 30-fold, enable local execution, and one-day transpile SQL across data platforms. “Today, we launched the dbt Fusion engine, a complete rewrite of dbt from the ground up,” he explains, framing the overhaul as essential for scale. The roadmap suggests dbt could slash warehouse costs and free teams from vendor lock-in, marking a bold shift from incremental tweaks to deep platform bets. (getdbt.com) Introducing Claude 4 | Anthropic (May 22 2025)Anthropic’s Claude 4 family—Opus 4 and Sonnet 4—promises state-of-the-art coding, extended tool use, and improved memory for multi-hour agent workflows. The post touts that “Claude Opus 4 is the world’s best coding model,” citing a 72.5 percent SWE-bench score and new parallel tool execution. By keeping prices steady and expanding availability across AWS, Google, and its own API, Anthropic aims to cement Claude as developers’ go-to frontier model. (anthropic.com) Scrapling | GitHubScrapling is an open-source Python library that claims stealthy, high-performance web scraping with automatic adaptation to site changes and anti-bot defenses. Its README highlights that “Scrapling is a high-performance, intelligent web scraping library for Python that automatically adapts to website changes.” With more than 5,000 stars and an April 2025 release, the project shows continuing demand for lightweight, developer-friendly scraping tools that outsmart detection. (github.com) Apple Researchers Just Released a Damning Paper That Pours Water on the Entire AI Industry | Futurism (Jun 09 2025)An Apple research team argues that leading “reasoning” language models plateau and even collapse on complex puzzles, challenging industry claims of true machine reasoning. Their study warns that “frontier [reasoning models] face a complete accuracy collapse beyond certain complexities,” calling current performance an “illusion of thinking.” The finding could temper expectations for next-gen LLMs and intensify scrutiny of benchmarking just as Apple readies its own AI features. (futurism.com) Bill Atkinson Dies From Cancer at 74 | Daring Fireball (Jun 07 2025)John Gruber reports that pioneering Macintosh programmer Bill Atkinson passed away at home on June 5 after battling pancreatic cancer. His family shared that “he was at home in Portola Valley in his bed, surrounded by family,” remembering him as “a remarkable person.” Gruber calls Atkinson perhaps “the most essential” coder on the original Mac team, noting his innovations in QuickDraw, MacPaint, and HyperCard still shape software today. (daringfireball.net)

    1h 9m

About

We talk about what's happening in Data & AI every week (well, most weeks). No buzzword bingo or marketing fluff—just honest conversations about the tech that's changing everything. Sometimes we go off on weird tangents about programming languages, tech/startup life, or whatever shiny object caught our attention. Whether you're knee-deep in code or just curious about AI, come hang out with us!

You Might Also Like