AI News Today

Article URL: https://www.youtube.com/watch?v=9nUmVktlwvA Comments URL: https://news.ycombinator.com/item?id=48985723 Points: 1 # Comments: 0

<p>Every team I've worked with hits the same argument about three months after shipping something on top of an LLM: <em>"Should we still be paying per token, or should we just self-host?"</em></p> <p>And it's always argued on vibes. Someone read a blog post. Someone else got burned by a GPU bill once. Nobody's actually done the math. So here's the math — or at least the shape of it.</p> <h2> The API wins right up until it doesn't </h2> <p>Paying an API has one killer advantage that people forget

<p>Right now, somewhere in your org, someone is pasting a stack trace, a customer record, or a chunk of the codebase into ChatGPT to get unstuck. You know it. I know it. The person doing it half-knows it's sketchy and is doing it anyway — because it <em>works</em>.</p> <p>That's shadow AI, and the reason it's so hard to stamp out is that it's not laziness or malice. It's people trying to do their jobs faster. Send the "don't use ChatGPT" email and you don't stop it — you just make it quieter.</p

The director role for the Center for AI Standards and Innovation (CAISI) has become a revolving door since David Sacks left his position as czar.

Sony Music Entertainment has filed another lawsuit against Udio, accusing the AI music generator of infringing the copyright of more than 30,000 of its songs, ranging from Elvis Presley's Hound Dog to Beyoncé's Say My Name, and Harry Styles' As It Was. The lawsuit, filed in a New York court on Monday, claims that this […]

A Harvard mathematician says he used Anthropic's Fable 5 to disprove a major open math problem. How impressed should we be?

How I use Claude Code to challenge assumptions and turn research into a living artifact Continue reading on UX Collective »

Article URL: https://github.com/Open-Ingress/OpenIngress Comments URL: https://news.ycombinator.com/item?id=48985431 Points: 5 # Comments: 9

Article URL: https://seasonedleadership.substack.com/p/the-messy-reality-of-ai-transformation Comments URL: https://news.ycombinator.com/item?id=48985288 Points: 1 # Comments: 0

<pre style='white-space:pre-wrap;width:81ex'>fix(ui): keep agent files readable when unfocused (#111963)</pre>

Hi HN, I'm Minjae, a solo developer. - About two years ago, I came up with an idea and worked on it as a side project for about three months, but I had to quit because I was overwhelmed by my main job. However, when I recently used a coding agent to configure the specifications and have the development done, the implementation speed was incredibly faster than it was two years ago. So, I focused on refining the specifications and rode that momentum to launch the service in just two weeks. - The r

Article URL: https://router-website-ramp.vercel.app/router Comments URL: https://news.ycombinator.com/item?id=48985105 Points: 1 # Comments: 2

Article URL: https://lkforge.com/blog/26-repos-in-29-days/ Comments URL: https://news.ycombinator.com/item?id=48985091 Points: 1 # Comments: 0

Article URL: https://github.com/Gysho/LLMrPro Comments URL: https://news.ycombinator.com/item?id=48985063 Points: 1 # Comments: 1

Alphabet, Google's parent company, is reportedly working on a new chip designed to make its Gemini models run much more efficiently.

Article URL: https://fredheiding.substack.com/p/ai-voice-phishing-performs-on-par Comments URL: https://news.ycombinator.com/item?id=48985008 Points: 1 # Comments: 0

Enterprise AI is facing an ROI paradox. While throwing more compute at the strongest foundation model works well in product experiments, the costs become unbearable when the product is deployed in production. A new paper from researchers at Writer provides a solution that is accessible to engineering teams. The study takes a systematic look at optimizing the different components of the orchestration layer that wraps around the foundation model, aka the AI harness. By optimizing the harness, the

Judge grants restraining order, saying merger "likely to violate antitrust laws."

Article URL: https://runtimewire.com/article/ramp-router-ai-model-gateway-llm-costs Comments URL: https://news.ycombinator.com/item?id=48984923 Points: 2 # Comments: 0

Article URL: https://www.bloomberg.com/news/features/2026-07-16/how-ai-could-shrink-income-tax-revenue-and-threaten-economies Comments URL: https://news.ycombinator.com/item?id=48984802 Points: 5 # Comments: 2

Article URL: https://budgetlab.yale.edu/research/how-potential-ai-futures-would-play-out-current-tax-system Comments URL: https://news.ycombinator.com/item?id=48984725 Points: 4 # Comments: 2

Article URL: https://rcarmo.github.io/projects/memento/ Comments URL: https://news.ycombinator.com/item?id=48984705 Points: 1 # Comments: 0

Under the new system, the protocol will take a looser, "stateless" approach to session IDs on the server side, similar to how most ordinary websites already work.

<pre style='white-space:pre-wrap;width:81ex'>fix(ui): show new task prompts immediately without duplicates (#111953) * fix(ui): stabilize new task prompt rendering * test(ui): cover prompt handoff integration * test(ui): initialize prompt handoff fixtures</pre>

There are a large, and increasing, number of LLM-written comments on HN. If you're one of the people running these bots I have two questions for you: 1. Why? I assume most of them are for advertising/astroturfing/propaganda purposes, but I'm really hoping some of you are doing interesting experimentation at least in exchange for making us suffer through the slop. 2. How often does your bot's comment pass muster? I notice many of the obvious ones downvoted heavily, but I assume I'm being fooled a

Looking for contributors! We have 7 contributors and We have 11 "good first issue" labels and 17 "help wanted" issues. All well-scoped with code examples. Project: Nobody governs the context input layer. Tools exist to govern what agents DO (Microsoft AGT), what agents SAY (Guardrails AI), and how agents COMMUNICATE (NeMo). But the data flowing INTO the context window — retriever results, tool outputs, agent messages — is completely ungoverned. That's what Provena fixes Comments URL: https://new

Article URL: https://aventineresearchinstitute.substack.com/p/can-math-as-we-know-it-survive-ai Comments URL: https://news.ycombinator.com/item?id=48984365 Points: 1 # Comments: 0

Every coding session in claude code/ codex starts blindly. It forgets the archiectural decisions you made in a session you ran yesterday/ a week ago or a month ago. We end up depending on maintaining a Claude.md file or diff markdown files that these agents rely on their memory features. I started building CogniKernel to solve this problem. Instead of relying on markdown memory files, CogniKernel observes the coding session itself. It captures decisions, constraints, architectural decisions, pro

Article URL: https://hn.is-ai-good-yet.com/?view=frontpage Comments URL: https://news.ycombinator.com/item?id=48984040 Points: 4 # Comments: 0

With the releases of Fable 5, gpt 5.6 sol, and Kimi K3 I was curious how these models would perform building vivid 3d worlds, especially a Chinese open-weight one like Kimi. I also threw in some budget models so you can see how drastic the quality differences are. Comments URL: https://news.ycombinator.com/item?id=48983978 Points: 2 # Comments: 0

Built this to measure the semantic delta between what a site declares and what an AI agent actually reconstructs. Free, no signup. I'm the author — happy to answer questions. Comments URL: https://news.ycombinator.com/item?id=48983850 Points: 1 # Comments: 0

Talk of banning Chinese-made open-weight LLMs reveals the challenge of turning AI into a business.

Used an automated LLM workflow to take primary source documents to create a cited wiki. Then requested specific analyses and artifacts to be created from the archive. This is the result. Comments URL: https://news.ycombinator.com/item?id=48983681 Points: 3 # Comments: 0

<p>I keep hearing anecdotes from people who used coding agents to reverse-engineer and automate devices in their homes.</p> <p>I think this is an interesting illustration of the impact of the reduced cost of writing code.</p> <p>Prior to agents, it was entirely possible to reverse-engineer home devices. The problem was the ROI - was it really worth all of that effort? More importantly, any experienced programmer knows that undocumented, unstable APIs like that may well change or break in the fut

The one-year-old startup aims to reinvent financial architecture for autonomous AI transactions.

Article URL: https://lake8.dev/journal/dalla-keyword-alla-semantica-en/ Comments URL: https://news.ycombinator.com/item?id=48983393 Points: 1 # Comments: 1

"Abstract thought is not an all-or-none ability that suddenly appeared in humans." ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

https://theburmese.xyz/ I run a lot of parallel coding agents, and "is it done yet?" meant hunting across terminals and tmux panes. So i built Burmese, it natively discovers the agent sessions already running on your machine - Claude Code, Codex, and other harnesses. It doesn't copy or fork them: it resumes the real conversation by its native session ID. Every session, project, and agent lands in one live control center, so instead of hunting across terminals you glance at one place and know exa

Chinese open-weight model GLM 5.2 happily obliged