Postman silently auto-installs AI plugins and attempts to share all API data
Commenters say Postman lost users with forced logins, data collection and AI features; some switched to Bruno.
30 stories
Get a notification every evening
You hear about the digest when it is ready. Turn it off any time.
Notifications are on
You will hear about each evening’s digest when it is ready.
To get notifications on iPhone or iPad, add this site to your Home Screen first: tap Share, then Add to Home Screen. Open it from there to turn notifications on.
Notifications are blocked in this browser. Allow them in site settings and reload.
New notification sign-ups are closed for now. You can follow along with RSS.
Notifications could not be turned on. Check your connection and try again.
Notifications could not be turned off and are still on. Check your connection and try again.
The day’s top discussions
Commenters say Postman lost users with forced logins, data collection and AI features; some switched to Bruno.
Discussion revolves around the post being the fifth major article on scaling Git and questions about why code volume grows exponentially.
One commenter says the post is misleading and the key result is not about improving Qwen's math; another says the article inspired them.
Commenters discuss using iptables/nftables as a temporary mitigation and whether Ubuntu's delay in patching is enough to write it off.
Commenters discuss whether it's possible to avoid OpenAI using Codex chats for training and Google's dark patterns in Gemini.
Commenters discuss OccupySaaS starting an anti-lock-in protest after layoffs and that in SaaS you pay for the service, not the software.
Commenters list the recommendations in the NIST guide and argue such guides are well-intentioned but mostly ineffective.
One commenter questions whether such interpretations are correct, while another notes the experiment only shows impressive results with the latest models.
A commenter says the notes omit an important detail and finds the effect of beta values in RLHF confusing.
Commenters discuss the Platonic Representation Hypothesis; one says it might just be dimensionality reduction, while another trusts JEPA models.
One commenter says they analyzed a similar attack and couldn't report it to Google; another identifies the partner as Videostrong Technology Co.
One commenter finds the tagline over the top, while another wants the tool benchmarked against oh-my-pi.
One commenter disagrees with the claim that grandmasters analyze the same number of moves as amateurs; another speculates about old video compression.
Commenters note that the best baking paper is silicone vegetable parchment, while wax paper poses a fire risk in the oven.
Commenters discuss that the tool helps query data across databases but the README is poorly written.
Commenters recall first electronics projects built in Altoids tins; Beaglebone boards were designed to fit these tins.
Commenters note all code was written without AI; another says raycasting is fascinating to them.
Commenters criticize the study for using outdated models and lacking an agentic approach.
Commenters argue that token rationing lowers productivity and that self-hosting is more trustworthy for data security.
Commenters say Opus 5.5 is a game changer for drafting and that they use various models to create 3D representations of famous landmarks.
GitBook's CTO reported that they took a large portion of services offline due to the security attack and that systems are progressively coming back online.
Commenters note the project is from 2015, maintenance has tapered off, and an active fork exists; they also ask about systems combining code-CAD and traditional CAD.
Commenters say open-weight models are capable but the article's statistics are inaccurate; one commenter notes DeepSeek 4.1 handled most coding tasks without problem.
One commenter says moving from classical DLRM to generative recommenders involves challenges such as cardinality, cold starts, and training cost, and that no tutorial exists.
Commenters say LLMs read weights the 'human way' rather than through Matrix-like symbology; one commenter finds the Pac-Man example a stretch.
A smaller, more technical community
One commenter says Determinate Nix is the corporate version, while Lix aims to fix ecosystem papercuts.
Commenters say the post is really well written.
Commenters discuss Rust sudo incompatibility; one says sudo-rs does not aim to cover every feature and sees no license problem.
The week’s most-starred new repositories