Bold Text Doesn’t Control LLM Attention. So What Does?
A group of researchers wanted a model to pay closer attention to a span a user had marked. They did not solve it by marking the span more heavily. They went into the attention layer 🔎That choice is the most direct evidence I have found about what typing asterisks into...
What a Near-Identical App Taught Me About Building in Public
What a near-identical app taught me about building in public. Four days after I opened public testing for my nutrition app, I checked the Play Store for something unrelated and found another app with the exact same name. Same core idea, too: you tell it what you’re...
AI Hype Is Everywhere. Here’s How to Think Clearly Through the Noise
Every day brings a new headline about AI. One model reportedly outperforms another. A new system writes code, creates images, discovers drugs, or reasons through complex problems. Another allegedly breaks through security controls, exposes confidential information, or...
Speaking Siemens S7comm: Protocol Mechanics and Security Boundaries
S7comm is one of those protocols that looks simple until you start following the conversation all the way down. At first, there is just a TCP connection to port 102, then TPKT appears, then COTP, then Siemens-specific connection parameters, then the S7comm session has...
How yasbd-lib Reached v1.0 After Stress-Testing 39 Languages
Before 0.12.0, yasbd-lib was a promise. 39 languages, spaCy and pysbd integrations, a CLI, a streaming API, a benchmark suite. Every one of those language profiles took real time to build because we wanted them to survive actual text, not just the sentences we wrote...
Sydney Sweeney Sent Novig’s Search Interest Soaring. Can It Last?
The past week has been quite interesting for the prediction market sector, with discussion around Novig’s “Just Sports” campaign featuring Sydney Sweeney taking centre stage.Did it hit the sweet spot? On visibility, the answer appears to be yes. On 8 September, Novig...
The Most Dangerous AI Answer Is the One That Quietly Fails
The dangerous AI answer is not the one that fails. It is the one that looks right, reads with total confidence, pastes cleanly into your report or your code, and then does nothing. No error. No warning. The number on the page is still wrong, and now you are convinced...
Why I Built an Open-Source Project Manager Where AI Can Actually Take Action
Why I built Planvio, an open-source project management platform that treats AI as an operating layer instead of another chat box For the last few years, we’ve been adding AI to almost everything. Project management software is no exception. Open a modern...
Bonsai-2-27B-Ternary-CRACK-GGUF: A 27B Model With Refusals Removed
OverviewBonsai-2-27B-Ternary-CRACK-GGUF is a 27-billion-parameter text-generation model from dealignai. It applies proprietary weight-level abliteration to PrismML’s ternary Bonsai 2 27B release, removing the refusal circuitry while retaining the base model’s...
AI Labs Warn of Extinction. Their Safety Policies Tell a Different Story
Sometime in the second week of September you read a quote from a man who had just quit his job at an AI company. It said the people building this technology believe it could kill us all by the end of the decade. You read it on a screenshot, or somebody read it to you,...
Inside Tencent’s Platform Strategy: Weixin, Cloud and AI
A company behind familiar services Weeks 1 to 3 took me from e-commerce to digital payments and WeChat. For Week 4, my question is how Tencent connects these services with its wider business. I examine its platform strategy and enterprise tools using company sources,...
You Don’t Have to Leave Claude Code to Use DeepSeek
You do not have to give up Claude Code or Codex CLI to put DeepSeek to work, and you do not have to run a proxy either. Each client has its own documented route to the same API. 🔌Claude Code -> direct, Anthropic-compatible endpoint Codex CLI -> official setup...
My Fishing App’s Tamper-Proof Ledger Failed Its First Real Audit
My fishing app gives points for verified catches. Points decide what you can see on the map. So the first question anyone sensible asks is: what stops the operator, me, from quietly editing the numbers? The fashionable answer is a blockchain. I used thirteen lines of...
Build Customizable, Reusable Project Templates With Node.js
At some point in your life you end up with a repo that is the starting point. It has the CI pipeline you like, your preferred linting and formatting configs and the Terraform layout that works like a charm. Then a new project shows up, you clone the base template and...
BigCommerce app breach spills Master of Malt customer data
Hackers have raided Master of Malt's customer database after a compromised ecommerce app gave them four days to help themselves to names, addresses, phone numbers, and email addresses. The online booze retailer began notifying customers this week after learning that...
Outerlimit Raises $16 Million to Stop Rogue AI Agents From Causing Harm
Emerging from stealth with $16 million in pre-seed funding, Outerlimit offers a decentralized authorization layer designed to discover, observe, and block harmful autonomous AI actions. The post Outerlimit Raises $16 Million to Stop Rogue AI Agents From Causing Harm...
Ratio CEO John Cho on Stablecoin FX: Asia’s Won-to-Rupiah Payments Without the Dollar
A Korean manufacturer paying an Indonesian supplier does not, in practice, buy rupiah with won. It buys dollars with won, then buys rupiah with dollars, because no bank offers a deep direct pair between the two. The invoice pays two foreign exchange spreads instead of...
The Navy’s top officer lays out his vision for laser weapons on every ship
This article is republished with permission from Laser Wars, a newsletter about military laser weapons and other futuristic defense technology. U.S. Navy Admiral Daryl Caudle is the 34th Chief of Naval Operations and a longtime student of directed energy who’s...
Abandoning Scientific Linux Was a Mistake
Article URL: https://blog.melashri.net/posts/scientific-linux-mistake/ Comments URL: https://news.ycombinator.com/item?id=49813286 Points: 7 # Comments: 1
5 science-backed strategies to be a better boss
Below, Caroline Webb shares five key insights from her new book, Leadership Intelligence: Science-Backed Strategies for Mastering 21 Everyday Management Challenges. Caroline is an executive coach and economist who transforms insights from behavioral science into...
BT Tower’s rooftop pool plan brings swimmers back down to earth
London's iconic BT Tower is set to get a rooftop swimming pool in plans submitted by the developer, but it may not be quite the view that readers expect. MCR Hotels has submitted its planning application to Camden Council more than two years after BT agreed to sell...
Netherlands bracing for potentially devastating US sanctions against the ICC
Article URL: https://apnews.com/article/icc-trump-sanctions-eu-israel-netherlands-2c1cc314732f920c2de59396d3b556f6 Comments URL: https://news.ycombinator.com/item?id=49813419 Points: 12 # Comments: 0
How to tell if people who look good in meetings are just faking it
Whether you work remotely or in an office (or a mix of both), meetings are an integral part of work, especially for knowledge workers. According to recent estimates, the average American professional will spend more than 14 years of their working lives in meetings. As...
The surprising downside of high-performing employees
High-performing employees are valued by organizational leaders for good reason. They’re reliable, adaptable, and willing to step in when something needs to get done. When responsibilities are unclear or processes break down, they figure it out. When a team is...
How employers can make good remote work policies before the next crisis
More than six years after the COVID-19 pandemic began, more than one in five of the U.S. workers who can work remotely do so either full-time or at least one day per week. Even though a new normal has set in, no single approach has worked for all employees. And the...
F5 Patches Critical BIG-IP APM Zero-Day Exploited for Unauthenticated RCE on OAuth Servers
Attackers are exploiting a critical flaw in F5 BIG-IP Access Policy Manager (APM) that lets them run code on a BIG-IP system without logging in, F5 says. The flaw, CVE-2026-94127, affects only systems in which APM serves as an OAuth authorization server, issuing...
Chinese Hackers Exploit Chrome-Windows Zero-Day Chain to Deploy CLEANGULP Malware
A Chinese threat actor codenamed UTA0565 has been observed exploiting the recently disclosed Google Chrome-Microsoft Windows exploit chain as zero-days through fake websites. The attacks, detected on September 3 and 4, 2026, involved the chaining of two...
Arista Urges Immediate Patching of Exploited VCO Zero-Day
Remote attackers could trigger the critical-severity flaw to access privileged internal functionality. The post Arista Urges Immediate Patching of Exploited VCO Zero-Day appeared first on SecurityWeek.
UK’s uncrewed experimental sub shows it can fire torpedoes
A British uncrewed submarine has test-fired a torpedo for the first time, demonstrating the ability of undersea drones to attack targets as well as perform reconnaissance and surveillance roles. The XV Excalibur, a 12 meter (40 ft) experimental uncrewed underwater...
Jev in 25 Lines of Python
Article URL: https://www.nobodywho.ai/posts/jev-in-25-lines/ Comments URL: https://news.ycombinator.com/item?id=49812769 Points: 12 # Comments: 4