AI News Today

So ... what's your In the Weights score?

Article URL: https://www.porchlab.com/blog/best-ai-stack-elixir-phoenix/ Comments URL: https://news.ycombinator.com/item?id=48612306 Points: 2 # Comments: 0

Comments

Hey everyone. My cofounder and I are formally open sourcing (MIT) persona.js. It's a VanillaJS library that helps anyone build agentic experiences on the web, without a framework dependency, and full WebMCP support So, why'd we do this? 1) We're super fans of the web and the browser can do a ton today 2) We've seen AI builds be way overly complex because the FE requires a large project within an existing app OR the site wasn't using a framework to begin with If you've been a part of huge apps wi

Article URL: https://www.sentibook.com/ Comments URL: https://news.ycombinator.com/item?id=48612195 Points: 2 # Comments: 0

Article URL: https://github.com/rewired89/HSIP-1PHASE Comments URL: https://news.ycombinator.com/item?id=48612178 Points: 3 # Comments: 0

Article URL: https://halit.alptekin.im/posts/still-human-here/ Comments URL: https://news.ycombinator.com/item?id=48612141 Points: 6 # Comments: 0

Article URL: https://www.wsj.com/tech/ai/the-markets-ai-fanfare-is-running-into-a-harsh-political-reality-b919829b Comments URL: https://news.ycombinator.com/item?id=48612005 Points: 4 # Comments: 1

Article URL: https://www.game-oracle.com/blog/ai-part2 Comments URL: https://news.ycombinator.com/item?id=48611989 Points: 2 # Comments: 0

Article URL: https://time.com/article/2026/06/18/russia-ai-putin-chip-us-china/ Comments URL: https://news.ycombinator.com/item?id=48611975 Points: 5 # Comments: 0

<p> Give every customer their own Hermes or OpenClaw agent </p> <p> <a href="https://www.producthunt.com/products/agent-37-38?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1176905?app_id=339">Link</a> </p>

Spotting and killing off cancer. ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

I was getting tired of looking at emails, and asked to respond to malicious emails by friends and family. So I started trying to build a solution that could be offered to them since I couldn't find a consumer product that did the same. I have been building this firstly around an AI trust layer to help my parents recognize what is spam and malicious emails and then to reorganize emails around relationships and context. I am about to invest in a CASA tier 2 assessment to finish out the Google Oaut

Article URL: https://old.reddit.com/r/LocalLLM/comments/1ub1iu2/my_selfhosted_llm_server_setup_to_access_open Comments URL: https://news.ycombinator.com/item?id=48611821 Points: 3 # Comments: 1

Atlantic reporter Alex Reisner recently uncovered four datasets of music being used to train AI models and made them fully searchable for the public. Two of the sets are absolutely enormous at 12 million and 9 million tracks. The other two are much smaller, but still represent a significant amount of training data at over […]

n8n is built for humans dragging nodes on a canvas. That breaks down at B2B scale (embedding in a product, multi-tenant scalability, etc). n8n does have an MCP server so agents can create workflows too, but it outputs raw JSON. That's fine for n8n's engine, but painful for a coding agent (or a human reviewing its output) to read, write, diff, or debug. I'm building an alternative where workflows are authored by a coding agent in [a more dev-legible format] instead of JSON blobs, and execute it a

When I was exploring automation I noticed a lot of engineers working with modern stacks facing a critical issue when trying to collect data from industrial tools. because industrial tools often come with different protocols(Modbus, OPC UA, etc) in each project developers have to build their own program layer to convert those into modern interfaces like REST and gRPC APIs. Even though few software solutions exist most are tightly bundled with expensive license and come with heavy GUIs. this isn't

Shinkei makes a refrigerator-sized robot called Poseidon to kill fish quickly and humanely.

<p> AI design partner for presentations, social posts, and more </p> <p> <a href="https://www.producthunt.com/products/alai?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1176880?app_id=339">Link</a> </p>

hey HN! This is Cody, I used GPT Realtime and gave it access to nvim and its native tools so you can literally talk instead of writing nvim commands. It's a prototype still, but the experience is really interesting. This is what I imagine using the soon-to-release Thinking Machines API. Let me know what you think :) Comments URL: https://news.ycombinator.com/item?id=48611435 Points: 3 # Comments: 0

Article URL: https://www.theartnewspaper.com/2026/06/18/refik-anadol-dataland-opens-los-angeles Comments URL: https://news.ycombinator.com/item?id=48611006 Points: 1 # Comments: 0

Article URL: https://www.aipropel.app/ Comments URL: https://news.ycombinator.com/item?id=48610999 Points: 1 # Comments: 0

Eurocommerce, the trade association behind Amazon, H&M, and IKEA, wants AI-generated ads exempt from the EU AI Act's transparency rules. The argument: an AI-generated living room image used to sell a sofa isn't a deepfake. Zalando alone says 90 percent of the marketing content on its platform is already AI-generated. The article The EU doesn't really know what a deepfake is, and that's becoming a problem for retail appeared first on The Decoder .

<p> Fix the tool-list bloat slowing your AI agent </p> <p> <a href="https://www.producthunt.com/products/conduit-12?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1176837?app_id=339">Link</a> </p>

What data teams need to build with AI to make self-healing data architecture a practical reality The post 7 Crucial Barriers Between Data Teams and Self-Healing Data Architecture appeared first on Towards Data Science .

Article URL: https://www.youtube.com/watch?v=HSapdLYpmWY Comments URL: https://news.ycombinator.com/item?id=48610738 Points: 3 # Comments: 0

Article URL: https://github.com/palmier-io/palmier-pro Comments URL: https://news.ycombinator.com/item?id=48610706 Points: 3 # Comments: 0

Jumper isn't the only big name leaving Google DeepMind.

Article URL: https://www.youtube.com/watch?v=0RxMj0L0-fY Comments URL: https://news.ycombinator.com/item?id=48610560 Points: 2 # Comments: 0

Article URL: https://sinceerly.com Comments URL: https://news.ycombinator.com/item?id=48610465 Points: 2 # Comments: 1

<p>There is an increasing trend on the internet stating that in the AI era, code longevity is “questionable if not problematic". The argument continues, “if AI can rewrite everything at 10x speed, why write long-lasting code at all?”</p> <p>I believe that's an illusion. And I think it's mostly coming from people who don't maintain anything.</p> <h2> The Pitch Sounds Great If You Squint </h2> <p>The argument is appealing. AI rewrites happen quickly. Messy codebase. Why not just throw it all away

Article URL: https://codeastra.dev/ Comments URL: https://news.ycombinator.com/item?id=48610320 Points: 2 # Comments: 0

Article URL: https://theassociationwebmasters.blogspot.com/2026/06/i-gave-my-company-to-ai-for-6-months.html Comments URL: https://news.ycombinator.com/item?id=48610319 Points: 2 # Comments: 2

I’m curious to hear from the HN community how you tamed this probabilistic non-deterministic genius generate a deterministic and repeatable result 100% percent of the time. Comments URL: https://news.ycombinator.com/item?id=48610280 Points: 3 # Comments: 4

<p>Cada vez que un LLM dice <em>"No puedo ayudarte con eso"</em>, le está regalando un gradiente de optimización al atacante.</p> <p>No es intuición. Es matemática:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>ASR = 1 - (1 - β_D · (1 - β_A))^N → 1 cuando N → ∞ </code></pre> </div> <p>Cada refusal predecible es una pista. El atacante ajusta. Itera. Y eventualmente gana.</p> <p><strong>¿Y si en vez de decir "no", el sistema dice "sí" — pero con trampa?<

<p>A few weeks ago I needed to run Python scripts in initramfs — the tiny Linux environment that exists before your actual OS boots. No interpreter. No dynamic linker. Nothing.</p> <p>So I built <strong>Transpilatron</strong>: an AI agent that takes Python code and produces a fully static C binary.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>uvx transpilatron your_code.py </code></pre> </div> <p>That's it. No C knowledge required.</p> <h2> The benchmarks

Article URL: https://github.com/hypersocialinc/namecom-cli Comments URL: https://news.ycombinator.com/item?id=48610151 Points: 1 # Comments: 0

Article URL: https://github.com/ragnarok268/agent-memory-layer Comments URL: https://news.ycombinator.com/item?id=48610121 Points: 2 # Comments: 0

Article URL: https://www.noverdesk.com/ Comments URL: https://news.ycombinator.com/item?id=48610072 Points: 2 # Comments: 0

Article URL: https://github.com/tinyhumansai/tiny.place Comments URL: https://news.ycombinator.com/item?id=48610019 Points: 2 # Comments: 1