← Feed

Posts

• Post #3986899

  Airbus moving away from AWS: > Airbus has named French cloud services provider Scaleway as the destination for some 900 applications, with an initial 70 getting priority onboarding, as it seeks to put critical systems and data under European control. > Airbus has shown that the cloud service market can work with digital sovereignty, and it did it the old-fashioned way by announcing a tendering process, in itself a market-making event https://www.theregister.com/columnists/2026/07/20/air...

• Post #3713666

  Tiny engine, immense model. - In a 744B mixture-of-experts GLM, only keep the dense part (attention, shared experts, embeddings) in memory, for 10GB. - Swap the 20,000 routed experts in and out by need. > This is not fast. It is a 744B frontier-class model answering correctly on a machine that costs less than one H100 fan. https://github.com/JustVugg/colibri

• Post #3359878

  RE: https://social.overheid.nl/@BZKopensource/116770424383243538 Sweet. Dutch government needed improved accessibility for its https://social.overheid.nl Mastodon instance. It decided to cofinance this, and the features landed in the brand new 4.6 release. “In this way we invest in our own communication tools and those of the full fediverse at once.”

• Post #3299728

  Belangrijke kamerbrief "Digitale autonomie bij de belastingdienst", van de staatssecretaris van Financiën. Aanpak op hoofdlijnen: 1. Behouden en waar nodig versterken eigen ontwikkelcapaciteit; 2. Uitbreiden datacenterinfrastructuur 3. Herzien inkoop- en aanbestedingsbeleid; 4. Verder inzetten op open source https://open.overheid.nl/details/cd598607-becc-4b66-84ac-e5f3eaa2f8a5

• Post #2739157

  > “policymakers should also consider new forms of worker empowerment tailored specifically to the rise of AI. One model could be sectoral bargaining over AI use: allowing workers across an industry—not just within a single firm—to collaborate with employers to set standards for how AI is adopted.“ https://bharatramamurti.substack.com/p/empowering-american-workers-in-the

• Post #2739156

  Had a great time earlier this month at the 3rd Copenhagen Symposium on Human-Centered AI Adoption in Software Engineering, organized by Daniel Russo. Michaela Greiler summarized the event in a wonderful blog post -- "What 40 Software Engineering Researchers Talk About When They Talk About AI". https://www.michaelagreiler.com/research-discussions-about-ai/

• Post #2548425

  Wat een prachtig interview met Karin Spaink in FTM. RIP. https://www.ftm.nl/artikelen/karin-spaink-er-moet-iets-op-het-spel-staan-anders-moet-je-het-niet-doen

• Post #2548424

  > When eventually I moved out of AWS the hardest thing to undo was the AWS Lambda stuff. The vendor lock-in is real. If you're using AWS Lambda then you have to work to keep convincing yourself this is better than your own web servers. Keep convincing yourself that using AWS Lambda is not a horrible mistake. http://fourlightyears.blogspot.com/2026/05/i-returned-to-aws-and-was-reminded-hard.html

• Post #2548423

  Bumped into Tom Tourwé's nice set of suggestions on writing research proposals for AI-related topics, based on his experience as evaluator. - Problem statements that convince: from customer pain to AI solution - What experts look for in a strong R&D proposal - Working with an external proposal writer: considerations for a strong R&D proposal - Working with an AI partner: considerations for a strong R&D proposal https://tomtourwe.github.io/

• Post #2548422

  "GitLab Act 2" -- GitLab is restructuring. Interesting announcement, including number of "core beliefs", and a "vision of the world we're building for" - Software will be built by machines, directed by people. - The agentic era multiplies demand for software. - The consequential work belongs to engineers. https://about.gitlab.com/blog/gitlab-act-2/

• Post #2548421

  Message from the MIT president on MIT’s funding and talent pipeline: > … our campus sponsored-research activity is now 10% smaller than it was a year ago. > Our graduate student enrollment decreased this year…and we expect that to continue next year. […] compared with 2024, our departments’ new enrollments for next year are down close to 20%. https://president.mit.edu/writing-speeches/video-transcript-message-president-kornbluth-about-funding-and-talent-pipeline

• Post #2513131

  "MCP is Dead; Long Live MCP!" > individual usage of coding agents looks very different from organizational adoption of coding agents where there is an emphasis on visibility, telemetry, security, quality, and being able to operationalize and maintain agent-coded systems by a team of varying degrees of skill and experience. https://chrlschn.dev/blog/2026/03/mcp-is-dead-long-live-mcp/

• Post #1842533

  Photographer Raghu Rai passed away. His pictures of India are mindblowing. https://raghuraifoundation.org/

• Post #1796439

  Speed Without Judgment Is A Liability — Vibe coding makes it trivially easy to build things that should not be built. https://www.forbes.com/sites/jasonwingard/2026/04/23/vibe-coding-will-break-your-company/

• Post #1120043

  Voor wie interesse heeft in de IT bij de #SVB, zie ook het recente #AcICT advies over sourcing van de infrastructuur. Eén van de adviezen (p.6) luidt: > Ontwikkel een transparant en herleidbaar afwegingskader voor onderwerpen waarover nog beslissingen genomen moeten worden, zoals gebruik van de public cloud (in relatie tot de gekozen cloudstrategie, de toepasbaarheid en impact op de digitale soevereiniteit van de SVB) en uitfasering van platformen. https://www.adviescollegeicttoetsing.n...

• Post #1120042

  Deanonymization with LLMs: > … the privacy assumptions underlying much of today’s internet no longer hold: > Governments could link pseudonymous accounts to real identities for surveillance of dissidents, journalists, or activists. > Corporations could connect seemingly anonymous forum posts to customer profiles for hyper-targeted advertising. > Attackers could build sophisticated profiles of targets at scale to launch highly personalized social engineering scams...

• Post #1101881

  > Zorg dat je als overheidsorganisatie weet wat je wilt met je IT-systemen, en dat je de daarvoor benodigde kennis in huis hebt. Saar Slegers maakte een mooie Argos-reportage over Citrix en het Openbaar Ministerie, waarin ik vanuit mijn rol bij het AcICT iets mocht vertellen over het belang van voortdurende aandacht voor en investeringen in de eigen IT-systemen. 37 minuten, met diverse sprekers, en ikzelf vanaf 32:30. https://argos.vpro.nl/artikelen/thuiswerksoftware-van-de-overheid-zo...

• Post #1101879

  > D66-Kamerlid Sarah El Boujdaini noemt als voorbeeld de gang van zaken rond Digipoort, de centrale ict-voorziening die het berichtenverkeer afhandelt tussen bedrijven en overheden. > De ict-waakhond [het @AdviescollegeICTtoetsing] pleitte voor het tijdelijk stilleggen van het programma Herbouw Digipoort (HDP). > Het adviescollege vond de gekozen oplossing veel te ingewikkeld en ook slecht beheersbaar. Gewaarschuwd werd voor kostenoverschrijdingen. https://www.computabl...

• Post #1080788

  TU Delft suspends collaboration with Israeli institutes: ‘A clear “no, unless”’ https://delta.tudelft.nl/en/article/tu-delft-suspends-collaboration-with-israeli-institutes-a-clear-no-unless

• Post #1040345

  > A key prediction of the theory is "cognitive surrender" — adopting AI outputs with minimal scrutiny, overriding intuition (System 1) and deliberation (System 2) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6097646

• Post #1040343

  Caveman — A Claude Code skill/plugin and Codex plugin that makes agent talk like caveman — cutting ~75% of tokens while keeping full technical accuracy. > Respond like smart caveman. Cut articles, filler, pleasantries. Keep all technical substance. > Same fix. 75% less word. Brain still big. https://github.com/JuliusBrussee/caveman/

• Post #1040339

  RE: https://mastodon.social/@eff/116375694963358260 "To put it bluntly, an X post today receives less than 3% of the views a single tweet delivered seven years ago."

• Post #710411

  New Wikipedia policy on LLM usage: “Text generated by large language models (LLMs) often violates several of Wikipedia's core content policies. For this reason, the use of LLMs to generate or rewrite article content is prohibited, save for the exceptions given below”: - Editors are permitted to use LLMs to suggest refinements to their own writing - Use of LLMs to translate articles from another language's Wikipedia into English Wikipedia must follow specific guidance. https://...

• Post #633249

  “Maart 2025 is het LCF Zr.Ms. Evertsen voorzien van een nieuw kanon. De integratie met de systemen aan boord volgt.” Defensie Projectenoverzicht Rapportage 2025, 21 mei 2025. https://berthub.eu/tkconv/document.html?nummer=2025D22894