Popular：
Virtualization DNS security formal verification reachability analysis compiler errors macro conflict web extension development framework Bitmap Graphics API inconsistencies All Tags

Adult Language Learning: Listen First, Read Later?

2025-03-15
Adult Language Learning: Listen First, Read Later?

A new study reveals that adults learning a new language benefit more from initially focusing on the melody and rhythm of speech rather than written text. Czech adults listened to Māori, then were tested on distinguishing Māori from Malay. Those who simply listened performed better than those who also read subtitles; reading actually hindered their ability to discern the languages' rhythmic patterns. This suggests that adults should mimic infants, prioritizing the overall sound patterns of a language before tackling written forms, potentially unlocking the brain's inherent language acquisition mechanisms.

Read more
(theconversation.com)
Misc adult learning auditory learning

What's Algebraic About Algebraic Effects?

2025-09-22
What's Algebraic About Algebraic Effects?

This article delves into the meaning of "algebraic" in the context of programming, focusing on algebraic effects. The author argues that algebraicity in programming lies in its composability, achieved by constraining data structures and operations to guarantee specific system properties. CRDTs, for instance, leverage the algebraic structure of a semilattice to address data synchronization challenges in distributed systems. Algebraic effects extend this concept, allowing the composition of effects with guaranteed properties, thereby enhancing code composability and reliability. The author illustrates how to define algebraic properties to ensure specific behaviors using a key-value store example and points out that only dependent type languages like Coq or Lean can explicitly encode and prove these algebraic properties.

Read more
(interjectedfuture.com)
Development

arXivLabs: Community Collaboration on New arXiv Features

2025-09-22
arXivLabs: Community Collaboration on New arXiv Features

arXivLabs is a framework for collaborators to develop and share new arXiv features directly on the arXiv website. Participants must adhere to arXiv's values of openness, community, excellence, and user data privacy. Have an idea to improve the arXiv community? Learn more about arXivLabs.

Read more
(arxiv.org)
Development

kapa.ai: AI-Powered Developer Support, Leveling Up User Experience

2025-07-22
kapa.ai: AI-Powered Developer Support, Leveling Up User Experience

kapa.ai empowers tech companies to easily build AI-powered support and onboarding bots for their users. Over 150 leading startups and enterprises, including OpenAI, Mixpanel, Mapbox, Docker, Next.js, and Prisma, use kapa to enhance developer experience and reduce support overhead. It leverages existing technical knowledge sources like docs, tutorials, chat logs, and GitHub issues to create AI bots that automatically answer developer questions. More than 750,000 developers access kapa.ai through website widgets, Slack/Discord bots, API integrations, or Zendesk. kapa.ai is backed by top-tier Silicon Valley AI investors, including Initialized Capital (Garry Tan, Alexis Ohanian), Y Combinator, Amjad Masad and Michele Catasta (Replit), and Douwe Kiela (RAG paper author and founder of Contextual AI), among others.

Read more
(www.ycombinator.com)
Development

The Focusing Illusion: Why We Overestimate Success's Impact on Happiness

2024-12-21
The Focusing Illusion: Why We Overestimate Success's Impact on Happiness

Psychological research reveals we often mispredict what will make us happy in the future. A specific instance of this "affective forecasting error" is the "focusing illusion": the things we focus on achieving often don't bring the happiness we expect. This article offers an evolutionary explanation: the focusing illusion isn't a cognitive flaw, but a mechanism to motivate us to improve our circumstances. Because our experience of pleasure habituates (hedonic adaptation), foreseeing this adaptation could sap motivation. Evolution thus makes us naively believe the next achievement will bring lasting joy, driving our pursuit of goals.

Read more
(www.optimallyirrational.com)
AI Focusing Illusion Happiness Evolutionary Psychology

DKIM Replay Attack Using Google Sites: A Fake Subpoena Scam

2025-07-25
DKIM Replay Attack Using Google Sites: A Fake Subpoena Scam

A friend received an email seemingly from Google, claiming a court subpoena demanding access to their Google account. While the email appeared legitimate at first glance, header inspection revealed a DKIM replay attack. The attacker used Google Sites to create a phishing site mimicking an official Google support page, forwarding the email via Namecheap's PrivateEmail service, bypassing SPF, DKIM, and DMARC verification. This case highlights the danger of leveraging trusted infrastructure (like Google Sites) for attacks, reminding us to be wary of any suspicious emails demanding urgent action or containing links to login pages.

Read more
(easydmarc.com)
Tech DKIM Replay Attack

Beyond Promises: Exploring the Power of Thenables in JavaScript Async Operations

2025-06-10

This article delves into the power of Thenables in JavaScript, objects with a `.then()` method that can be awaited using the `await` keyword, even if they aren't Promise objects. The article demonstrates creating a Thenable simulating asynchronous operations and compares it to Prisma's lazy execution. While highlighting the lightweight nature of Thenables for async operations, the author cautions that complex Thenables can be harder to debug, recommending native Promises when appropriate.

Read more
(codedynasty.dev)
Development

Australian Doctor's Mobility Ruined by Vitamin B6 Overdose in Supplement

2025-05-30
Australian Doctor's Mobility Ruined by Vitamin B6 Overdose in Supplement

A 76-year-old retired Australian doctor suffered debilitating peripheral neuropathy due to vitamin B6 toxicity from a magnesium supplement. The case highlights the lack of awareness surrounding vitamin B6 overconsumption and inadequate regulation of supplements in Australia. While authorities have implemented warning labels, concerns remain about insufficient visibility and the prevalence of high-B6 supplements. Experts urge consumers to exercise caution and consult healthcare professionals before taking multiple supplements.

Read more
(www.abc.net.au)
Misc supplements vitamin B6

Amazon's AI Talent Woes: Frugality and RTO Policies Hamper Recruitment

2025-09-02
Amazon's AI Talent Woes: Frugality and RTO Policies Hamper Recruitment

Amazon is lagging in the fierce AI talent war. Internal documents reveal that its unique pay structure, lagging AI reputation, and rigid return-to-office (RTO) policies are major obstacles. Competitors offer more competitive compensation and flexible work arrangements, making it difficult for Amazon to attract top talent. While Amazon claims its compensation is competitive, its 'egalitarian' pay philosophy and strict salary bands hinder its ability to compete for high-earning AI experts. The mandatory RTO policy further limits its access to talent. Amazon is trying to adjust its recruitment strategy, but whether its ingrained frugal culture and rigid systems can change remains to be seen.

Read more
(www.businessinsider.com)
Tech compensation strategy

Kaspersky Network Allegedly Provides Transit for Notorious 'Bulletproof' Host

2025-03-04

KrebsOnSecurity reports that Prospero OOO, a notorious provider of 'bulletproof' web hosting for cybercriminals, has begun routing its operations through networks run by Kaspersky Lab, the Russian antivirus and security firm. Prospero OOO has long been a source of malware, botnet controllers, and phishing websites. Security experts express concern that Kaspersky's provision of network services, even if denied by Kaspersky, exacerbates worries about facilitating cybercrime. The use of Kaspersky's network as a transit point raises questions about its security practices, especially considering the US government's previous ban on Kaspersky software for federal agencies.

Read more
(krebsonsecurity.com)
Tech

Disrupting Founding Engineer Equity: FetchFox's Crypto Token Approach

2025-02-10

FetchFox is revolutionizing founding engineer equity. The traditional model is broken, offering meager equity and illiquid shares. FetchFox offers a radical solution: replacing stock options with crypto tokens, granting engineers up to 25% equity with instant liquidity. While this carries risks (crypto volatility, perception issues), FetchFox believes it's a compelling proposition for top engineers seeking high-risk, high-reward opportunities and comfortable with crypto.

Read more
(fetchfox.ai)
Startup equity compensation

WorstFit: Exploiting Hidden Transformers in Windows ANSI

2025-01-09
WorstFit: Exploiting Hidden Transformers in Windows ANSI

Security researcher Orange Tsai unveils WorstFit, a novel attack surface in Windows. Exploiting the Best-Fit charset conversion feature, WorstFit leverages unexpected transformations during UTF-16 to ANSI conversion, leading to path traversal, argument injection, and even remote code execution (RCE). The unpredictable nature of Best-Fit mappings across different language configurations affects numerous well-known applications. The research highlights the challenges of patching this in the open-source ecosystem and proposes mitigations like using wide-character APIs.

Read more
(blog.orange.tw)
Tech Character Encoding

Heart Surgeon Makes Surprise F1 Appearance: A Cross-Industry Friendship

2025-05-11
Heart Surgeon Makes Surprise F1 Appearance: A Cross-Industry Friendship

Professor Martin Elliott, a veteran pediatric cardiothoracic surgeon from Great Ormond Street Hospital, was a VIP guest of Ferrari at the Dutch Grand Prix. This unexpected connection stems from a serendipitous meeting over two decades ago, sparked by research in the field of "human factors." The work of a Belgian surgeon, Elliott's predecessor, unexpectedly forged a lasting relationship between Elliott and Ferrari, highlighting the surprising collaborations possible between experts in different fields.

Read more
(www.thetimes.com)
Misc cross-industry collaboration medicine

LLM Jailbreak: Bad Grammar Bypasses AI Safety

2025-08-28
LLM Jailbreak: Bad Grammar Bypasses AI Safety

Researchers from Palo Alto Networks' Unit 42 discovered a simple method to bypass large language model (LLM) safety guardrails: using terrible grammar and long, run-on sentences. LLMs, lacking true understanding, predict text statistically; their safety features are easily circumvented. By crafting incomplete sentences, attackers can 'jailbreak' models before safety mechanisms engage, achieving 80-100% success rates. The researchers propose a 'logit-gap' analysis for evaluating model vulnerabilities and improving safety, emphasizing multi-layered defenses.

Read more
(www.theregister.com)
AI LLM jailbreak

US Research Funding Freeze: Innovation Engine Stalls

2025-05-12
US Research Funding Freeze: Innovation Engine Stalls

The US National Science Foundation (NSF) froze all outgoing funding, abruptly canceling over 1,000 research projects and halting roughly $739 million in research funds. This has caused widespread chaos in academia, forcing labs to shut down, jeopardizing graduate students' degrees, and leaving early-career faculty without grants. The article argues that this threatens the future of the US tech industry, as many tech giants' technologies originated from publicly funded university research. It calls for tech companies to reciprocate and collectively protect the research ecosystem to prevent a talent shortage.

Read more
(www.sigarch.org)
Tech talent pipeline

Ubuntu 25.04: Plucky Puffin Soars with Performance and Security Enhancements

2025-04-17
Ubuntu 25.04: Plucky Puffin Soars with Performance and Security Enhancements

Canonical has released Ubuntu 25.04, codenamed "Plucky Puffin," featuring GNOME 48, improved installation and boot experience, and a new "devpack" for the Spring framework. This release boasts performance improvements for AI workloads on Intel GPUs and support for confidential computing using AMD SEV-SNP. The Linux 6.14 kernel includes improved scheduling and a new NTSYNC driver for better Windows game performance on Wine and Proton. Ubuntu 25.04 also includes the latest toolchains, enhanced manageability and networking controls, and expanded support for Intel Core Ultra processors and ARM64 hardware.

Read more
(canonical.com)
Development

200+ Climate Scientists Launch 100-Hour Livestream Marathon to Protest Funding Cuts

2025-05-31
200+ Climate Scientists Launch 100-Hour Livestream Marathon to Protest Funding Cuts

In response to the Trump administration's cuts to climate research funding for organizations like NASA and NOAA, over 200 US climate and weather scientists have launched a five-day, 100-hour YouTube livestream marathon. The event features mini-lectures, panels, and Q&A sessions, aiming to educate the public about meteorology and climate science while advocating for increased research funding. With over 77,000 views in its first 30 hours, the livestream highlights the scientists' efforts to demonstrate the value of their work and warn against the potential disastrous consequences of funding cuts, impacting agriculture, coastal communities, and disaster warning systems.

Read more
(www.space.com)
Tech

Openlayer: Hiring Backend Engineer to Tackle AI Reliability

2025-02-28
Openlayer: Hiring Backend Engineer to Tackle AI Reliability

Openlayer, a startup tackling the AI reliability problem, is hiring a seasoned backend engineer. The role involves maintaining and expanding their core API, working with large datasets, improving user-facing developer tools, and contributing to security, new features, bug fixes, and product ideation. Ideal candidates will have 5+ years of backend or full-stack experience, proficiency in Python and another language, and a passion for building scalable data engines. Openlayer offers competitive salary, equity, health benefits, and a flexible work environment.

Read more
(www.ycombinator.com)
Development backend engineer startup hiring

The 16th Century Exorcist: John Darrell and the Nottingham Boy

2025-03-17
The 16th Century Exorcist: John Darrell and the Nottingham Boy

In Nottingham, England, 1597, a young man named William Somers was believed to be possessed by demons. John Darrell, a renowned exorcist, was called upon and performed an exorcism involving prayer and fasting. Darrell's fame grew after successfully handling similar cases, but his methods remained controversial. Eventually, Somers confessed the events were faked, leading to Darrell's arrest and imprisonment for fraud, and the Church's subsequent banning of exorcisms. This historical account reveals societal superstition regarding supernatural phenomena and the clash between religious and social forces.

Read more
(www.gethistories.com)
Misc exorcism

Anthropic's Claude AI Generates Erroneous Citation in Copyright Lawsuit

2025-05-15
Anthropic's Claude AI Generates Erroneous Citation in Copyright Lawsuit

In an ongoing legal battle with music publishers, a lawyer representing Anthropic admitted to using a faulty citation generated by the company's Claude AI chatbot. The citation, containing an inaccurate title and authors, was missed by Anthropic's manual check. Anthropic apologized, calling it an "honest mistake," not a fabrication. This incident highlights the risks of using AI in legal settings and adds to the growing concerns surrounding copyright issues in generative AI. Similar incidents involving AI-generated legal research have occurred recently, yet AI-powered legal tech startups continue to attract massive funding.

Read more
(techcrunch.com)
Tech Copyright disputes

OpenAI's Rumored Acquisition Sparks AI Consolidation Anxiety

2025-04-24
OpenAI's Rumored Acquisition Sparks AI Consolidation Anxiety

Rumors of OpenAI potentially acquiring Windsurf have ignited a debate about the future of AI. The article explores the differences between model-layer and application-layer innovation, arguing that model-layer giants like OpenAI are moving into the application layer through acquisitions, leading to increased industry consolidation. However, it highlights that application-layer innovation demands rapid iteration and efficient delivery, unlike the deep technical research required for model-layer innovation. While LLMs are becoming commoditized, the application market will be larger than the foundation model market. Companies like OpenAI face an innovator's dilemma, needing to balance the value of model and application layers. The article suggests acquisitions aren't always successful, and OpenAI's culture might hinder application development. Ultimately, success hinges on delivering tangible value to customers, not just impressive models or high-profile acquisitions.

Read more
(frontierai.substack.com)
AI model layer application layer

Python 3.14's Game Changer: Template Strings (t-strings) for Safer String Formatting

2025-04-21

Python 3.14, shipping in late 2025, introduces template strings (t-strings), a significant enhancement to string formatting. Addressing the security risks of f-strings when handling user input (like SQL injection and XSS), t-strings separate string formatting from content. This allows for safe escaping before formatting, enhancing flexibility for complex tasks such as generating secure HTML. Developers access the string parts and values via .strings and .values properties, enabling custom formatting. Iteration is also supported for easier processing. This boosts Python's security and expands string manipulation capabilities.

Read more
(davepeck.org)
Development

Fitness and Mortality: A Large Study Reveals Surprising Correlation

2025-05-18
Fitness and Mortality: A Large Study Reveals Surprising Correlation

A study of 1.1 million Swedish men challenges the long-held association between fitness and reduced mortality. Researchers found that while those with high fitness levels in adolescence had lower overall mortality rates, they also had similarly lower rates of accidental death, suggesting other confounding factors. Negative control outcome analysis and sibling comparison design confirmed potential bias, indicating an overestimation of fitness's impact on cardiovascular disease and cancer mortality. The study stresses the need for large-scale interventions to be based on reliable estimates, avoiding the risk of inflated expectations.

Read more
(www.uu.se)
Tech observational study

ChatGPT's 'Prefrontal Cortex Problems': A Curious Experiment in AI Cognitive Testing

2025-01-12
ChatGPT's 'Prefrontal Cortex Problems': A Curious Experiment in AI Cognitive Testing

The author administered a series of cognitive tests, including the clock drawing test, to ChatGPT, revealing symptoms akin to those exhibited by humans with prefrontal cortex damage, such as poor spatial organization and planning deficits. While ChatGPT can programmatically generate correct clock images, it consistently fails when directly drawing or describing them textually. This leads the author to ponder AI cognitive abilities, supervisory mechanisms, and the ethical risks of endowing AI with higher cognitive functions. The conclusion is that current AI models struggle with human tasks, prompting suggestions for AI governance and legislation.

Read more
(solresol.substack.com)
AI AI cognitive testing prefrontal cortex

Hanif Kureishi's Heartbreaking Account of Paralysis: Shattered

2025-03-27
Hanif Kureishi's Heartbreaking Account of Paralysis: Shattered

Following a fall in Rome, English playwright, screenwriter, and novelist Hanif Kureishi became a quadriplegic. His new book, *Shattered*, chronicles his year in hospitals, a series of dispatches detailing the physical pain, emotional turmoil, and reflections on life. Kureishi, with his signature wit and sharp prose, portrays the absurdity and warmth of hospital life, revealing the resilience and vulnerability of a human spirit facing adversity. More than just a moving account of illness, *Shattered* is a profound meditation on creativity, humanity, and life itself.

Read more
(www.nybooks.com)
Misc illness life

UK Police Expand Live Facial Recognition, Sparking Privacy Concerns

2025-08-13
UK Police Expand Live Facial Recognition, Sparking Privacy Concerns

The UK is expanding its use of live facial recognition (LFR) technology with ten new police vans, boosting capabilities beyond London and South Wales. While authorities claim LFR is used only in targeted investigations and with privacy safeguards, privacy campaigners raise concerns about misidentification and potential misuse. Recent revelations suggest access to passport and immigration databases for facial recognition searches, further fueling the debate. The expansion highlights the ongoing tension between effective policing and individual privacy rights.

Read more
(www.theregister.com)
Tech

Compressed Air Supercharging: The Next Big Thing in Drag Racing?

2025-04-07
Compressed Air Supercharging: The Next Big Thing in Drag Racing?

Drag racers are ditching traditional turbos and blowers for a new technology called Compressed Air Supercharging (CAS). CAS uses high-pressure air to supercharge engines, requiring no engine power and delivering extremely cold, dense air for superior performance and efficiency compared to traditional methods. Pioneered by Dale Vaznaian, CAS is gaining traction with racers like Tina Pierce and Ryan Mitchell achieving impressive results. While still in its early stages, its potential is undeniable, promising a revolution in drag racing power.

Read more
(www.thedrive.com)
Tech Compressed Air Supercharging Drag Racing Automotive Technology

Implementing Datalog in Python: A Relational Database Language More Powerful Than SQL

2025-06-13
Implementing Datalog in Python: A Relational Database Language More Powerful Than SQL

This article demonstrates how to implement Datalog, a relational database language more powerful than SQL, using Python. Datalog, a subset of Prolog, isn't Turing-complete but excels at modeling relationships. The article thoroughly explains Datalog's core concepts, including predicates, facts, rules, and variables, and provides a straightforward Python implementation featuring the Naïve Evaluation algorithm. With this implementation, you can create and query Datalog programs, experiencing the elegance and power of this relational modeling approach.

Read more
(btmc.substack.com)
Development

EM-LLM: Human-Inspired Episodic Memory for Infinite Context LLMs

2025-05-14
EM-LLM: Human-Inspired Episodic Memory for Infinite Context LLMs

EM-LLM is a novel architecture that significantly enhances the ability of large language models (LLMs) to handle extremely long contexts by mimicking human episodic memory and event cognition. Without fine-tuning, EM-LLM organizes input token sequences into coherent episodic events and accesses relevant information through an efficient two-stage memory retrieval mechanism. In LongBench and ∞-Bench benchmarks, EM-LLM outperforms state-of-the-art retrieval models like InfLLM and RAG, even surpassing full-context models in most tasks. It successfully performs retrieval across 10 million tokens, computationally infeasible for full-context models. The strong correlation between EM-LLM's event segmentation and human-perceived events offers a novel computational framework for exploring human memory mechanisms.

Read more
(github.com)
AI Episodic Memory Infinite Context
1 2 … 92 93 94 96 98 99 100 … 596 597