Hacker News

Comments URL: https://news.ycombinator.com/item?id=47208481

Points: 2

# Comments: 1

Built a crisis monitoring dashboard for the current Middle East situation using zero paid APIs. News from RSS feeds (Al Jazeera, BBC, Reuters, AP), market data from TradingView embeds, live TV streams, and a map with conflict hotspots. Source: https://github.com/jayyvk/situationtracker

Comments URL: https://news.ycombinator.com/item?id=47208472

Points: 2

# Comments: 0

1. Specs and plans are source code: Specs and plans live in git alongside source code, not in chat history. A new agent reads arch.md for the big picture, then its specific spec. You always know why something was built.

2. Three models review every phase: Claude, Gemini, and Codex catch almost entirely different bugs. No single model found more than 55% of issues. If you only review with the model that wrote the code, you're missing half the bugs. 20 bugs caught before shipping. Claude Code found 5 bugs, Gemini and Codex caught another 15, including a severe security issue Claude missed.

3. Enforce the process, don't suggest it. A state machine forces Spec → Plan → Implement → Review → PR. The AI can't skip steps. Tests must pass before advancing. AIs don't stick to the plan by themselves, you need rails.

4. Annotate, don't edit. Most of the work is writing specs and reviews that guide the code, not hacking at files in an open-ended chat.

5. Agents coordinate agents. An architect agent spawns builder agents into isolated git worktrees. You direct the architect; it directs the builders. They message each other async.

6. Manage the whole lifecycle. Most AI tools help you write code faster — maybe 30% of the job. The other 70% is planning how, reviewing, integrating, deployment scripts, managing staging vs prod. Have AI run the whole pipeline from spec to PR and beyond.

Overall result: One engineer able to produce what a team of 3-4 would usually do. Measured 1.2 points better code on a 10 point scale vs claude code. Downsides: takes a lot longer, much more token usage, but still reasonable at $1.60 per PR.

We open sourced it: https://github.com/cluesmith/codev More details and raw results: https://cluesmith.com/blog/a-tour-of-codevos/

Comments URL: https://news.ycombinator.com/item?id=47208471

Points: 2

# Comments: 1

Article URL: https://cybercanon.org/project-zero-trust-a-story-about-a-strategy-for-aligning-security-and-the-business/

Comments URL: https://news.ycombinator.com/item?id=47208110

Points: 1

# Comments: 0

Article URL: https://alexchao.substack.com/p/the-hard-truth-about-ai-agents-what

Comments URL: https://news.ycombinator.com/item?id=47208103

Points: 1

# Comments: 1

Their understanding of the exoskeleton, the exo-wings, and what those can do is entirely different. And it is not just a matter of "their understanding." The actual capabilities are genuinely different, the relevant functionality is genuinely different, and the ways of mining chat pearls are genuinely different too.

Strangely enough, I tend to place the writer in the manager category - he effectively manages the actions of his characters and strives to produce certain impressions on his superiors, meaning his audience and readers.

The programmer ideally wants a precise formal solution that can be quickly and cheaply verified in debugging.

The manager needs to improve something in his system, and life will be the judge of that - slowly, tediously, expensively, and often with ambiguous interpretations.

The student needs to go deep into knowledge, and the scientist needs to go deep into knowledge and also do the manager's job - searching for ways to apply and extend that knowledge.

This is of course a very incomplete list of the meaningful variations. But it will be enough for you to start understanding that there are very different worlds in the use of AI - and this is true even within a single profession, from junior all the way up to investor, senior, architect, and so on. Programmers who build AI or build agents on top of it often have no idea about these other possibilities. They often don't even know the names of the most important fields of knowledge required for that other kind of work with AI.

Here you will find notes on these topics, fragments of various files that were meant to become secret but somehow never did: https://zenodo.org/records/18824868 "These are notes that are not obliged to be true, let alone universally true."

Comments URL: https://news.ycombinator.com/item?id=47208089

Points: 1

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/Chief_Mouser_to_the_Cabinet_Office

Comments URL: https://news.ycombinator.com/item?id=47208083

Points: 1

# Comments: 0

Article URL: https://interpreted.art/art/MlLrLHxRKZHdQsGkv3RT

Comments URL: https://news.ycombinator.com/item?id=47208075

Points: 1

# Comments: 0

country's most valuable resource. If data is a resource, then perhaps we need a sovereign wealth fund for it. Taxing data extractors cannot, however, be a get-out-of-jail-free card that allows them to run roughshod over individual privacy or civil liberty. For users of platform technology, transparency could be increased with "opt-in" provisions that allow them more control over how their data is used (as is the case with the EU's General Data Protection Regulation, and the even tougher proposals in California). The "opt-in" language should be clear and simple, with the burden of proof for violations on companies rather than individuals. Big Tech companies should also be required to keep audit logs of the data they feed into their algorithms, and be prepared to explain their algorithms to the public. "A recurring pattern has developed," says Frank Pasquale at the University of Maryland, "in which some entity complains about a major Internet company's practices, the company claims that its critics don't understand how its algorithms sort and rank content, and befuddled onlookers are left to sift through rival stories in the press." Companies should be prepared to make themselves open to algorithmic audits, as suggested by mathematician and Big Tech critic Cathy O'Neil, in case of complaints or concerns about algorithmic bias that could allow for discrimination in the workplace, healthcare, education, and so on. 7 Individuals should also have their digital rights legalized. Former Wired editor John Battelle has proposed a digital bill of rights that would assign possession of data to its true owner, which is, of course, the user and generator of that data, not the company that made off with it. He believes this notion should be so central that it should be enshrined as an amendment to the Constitution. As the Europeans have put it, people should also have a "right to be forgotten," in which companies must delete any data held on individuals should they wish it Two million Europeans have already made the choice to opt out. Finally, I would like to see a digital consumer protection bureau, with tough rules around discrimination by algorithms, and a system for ensuring that individuals can access and understand how their personal data is being used, as we can with credit scores today. All of this relates to the need for more transparency and simplicity in the discussion about Big Tech. Complexity (or the illusion of it) is too often used to avoid legitimate public interest questions, such as how propagandists get their messages across, or how users are tracked and valued. Companies should help us understand by opening the black box of their algorithms. This needn't be a competitive disadvantage; research has shown that it is the amount of data plugged into an algorithm, rather than the cleverness of the algorithm itself,

Comments URL: https://news.ycombinator.com/item?id=47208044

Points: 1

# Comments: 4

Article URL: https://news.artnet.com/art-world/church-fresco-giorgia-meloni-2742134

Comments URL: https://news.ycombinator.com/item?id=47208035

Points: 1

# Comments: 0

When building small SaaS tools, I kept running into the same problem: I’d design beautiful pages… Then completely rethink navigation later. So I flipped the workflow. Now I design the entire product structure first as a visual sitemap. Not a text list. Not a static diagram. A drag-and-drop map of every page and connection. What changed: I cut unnecessary pages before writing code Navigation debates disappeared Feature creep became obvious User flows felt cleaner It felt less like “designing pages” and more like designing a system. Because I couldn’t find a lightweight tool that felt simple enough, I built this workflow into my own product —

EPIC ( https://no-edit.lovable.app )

It includes a visual sitemap builder designed specifically for small SaaS founders who want to think in structure before UI. I’m not sure why more design tools don’t prioritize structure thinking upfront. Curious — Do you plan product structure visually first, or jump straight into UI?

Comments URL: https://news.ycombinator.com/item?id=47208033

Points: 3

# Comments: 5

Article URL: https://www.youtube.com/watch?v=mLiXIQjQ5CU

Comments URL: https://news.ycombinator.com/item?id=47208022

Points: 1

# Comments: 1

Last week I posted hmem here and got some good feedback. I've been heads-down since then and v2 is out.

Quick recap of the idea: AI coding agents forget everything between sessions. Worse, if you switch machines or tools, even in-session memory is gone. Hmem fixes this with a 5-level hierarchy — agents load only L1 summaries on startup (~20 tokens), then drill deeper on demand. Like how you remember "I was in Paris once" before you recall the specific café.

What's new in v2: The tree structure is now properly addressable. Every node gets a compound ID (L0003.2.1), so you can update or append to any branch without touching siblings. update_memory and append_memory work in-place — no delete-and-recreate.

Obsolete entries are never deleted, just archived. They stay searchable and teach future agents what not to do. A summary line shows what's hidden.

Access-count promotion with logarithmic age decay. Frequently-used entries surface automatically — but newer entries aren't buried just because older ones have more history.

Session cache with Fibonacci decay. Bulk reads suppress already-seen entries so you don't get the same context dumped every call. Two modes: discover (newest-heavy, good for session start) and essentials (importance-heavy, kicks in after context compression).

TUI viewer for browsing .hmem files — mirrors exactly what the agent sees at session start, including all markers and scoring.

Curator role — a dedicated agent that runs periodically, audits all memory files, merges duplicates, marks stale entries, prunes low-value content. Also accesible via skill "hmem-self-curate". Still MIT, still npx hmem-mcp init. GitHub: https://github.com/Bumblebiber/hmem

Comments URL: https://news.ycombinator.com/item?id=47208019

Points: 1

# Comments: 0

Article URL: https://www.washingtonpost.com/technology/2026/02/23/ai-economic-growth-gdp-mirage/

Comments URL: https://news.ycombinator.com/item?id=47207971

Points: 1

# Comments: 0

Article URL: https://github.com/Vinix24/vnx-orchestration

Comments URL: https://news.ycombinator.com/item?id=47207959

Points: 1

# Comments: 1

Hey HN, I built PlagiarismRemover.AI, a free tool that rewrites content to make it unique while preserving the original meaning. It uses NLP and deep learning to restructure sentences rather than just swapping synonyms. No signup required, no data stored. Would love feedback from the community.

Comments URL: https://news.ycombinator.com/item?id=47207937

Points: 2

# Comments: 0

Article URL: https://davide.im/posts/multi-agent-system/

Comments URL: https://news.ycombinator.com/item?id=47207928

Points: 2

# Comments: 0

I do a lot camera calibration.

As my need for accuracy increased, I started using mrcal, a great camera calibration tool. But I found it didn't fit my workflow well in terms of workflow and as a software dependency.

What I wanted from a calibration library was:

Accuracy (obviously)

Trivial dependency, both calibration time and runtime

Simple and notebook-friendly python API

Support for both OpenCV and spline distortion models

Great built-in analysis tools

Board warp estimation

Outlier filtering

So that's what I built!

Would love feedback from other robotics / computer vision folks.

Comments URL: https://news.ycombinator.com/item?id=47207907

Points: 1

# Comments: 0

Article URL: https://www.sciencedirect.com/science/article/abs/pii/S1878788626000159

Comments URL: https://news.ycombinator.com/item?id=47207900

Points: 1

# Comments: 0

Hi HN — I’m an indie developer. I built Nano Banana 2 Playground, a simple web demo for generating images from text and transforming existing images (image-to-image).

My goal was a fast, minimal UI that lets you iterate on prompts quickly and compare results without extra friction.

I’d love feedback on:

Prompt UX (what feels missing/confusing)

Which controls/presets are actually useful

How to communicate limitations and expected results clearly

Thanks for taking a look — happy to answer questions and improve it based on your suggestions.

Comments URL: https://news.ycombinator.com/item?id=47207881

Points: 1

# Comments: 0