February 12, 2026

51 articles from this day

Building Hybrid Search for RAG: Combining pgvector and Full-Text Search with Reciprocal Rank Fusion

Combining pgvector and full-text search with Reciprocal Rank Fusion significantly improves retrieval precision in RAG systems. This approach achieves an 84% retrieval precision by merging results from both methods, all within a single PostgreSQL database, eliminating the need for additional infrastructure.

Dev.to AI-Assisted Coding ai python postgres

z.ai's open source GLM-5 achieves record low hallucination rate and leverages new RL 'slime' technique

GLM-5, an open-source large language model by z.ai, achieves a record-low hallucination rate and features a novel RL technique called 'slime' for efficient training. With 744B parameters and competitive pricing, it positions itself as a powerful tool for enterprises looking to enhance their AI capabilities.

VentureBeat Building AI Products technology

MIT's new fine-tuning method lets LLMs learn new skills without losing old ones

MIT's self-distillation fine-tuning (SDFT) method enables large language models to learn new skills without losing previous knowledge, addressing the challenge of catastrophic forgetting. This technique allows enterprises to maintain a single model for multiple tasks, reducing costs and improving adaptability in dynamic environments.

VentureBeat Building AI Products orchestration

Lessons from Zig

Zig's programming language philosophy advocates for a minimal standard library, emphasizing the removal of components that do not justify their maintenance costs. This approach contrasts with C++'s perpetual obligations, suggesting that C++ should adopt similar principles to enhance its standard library's efficiency and relevance.

Hacker News Building AI Products building-ai-products

Anthropic Safety Researcher Quits, Warning 'World is in Peril'

A safety researcher from Anthropic has resigned, warning that advancements in AI pose significant risks to global safety, including bioterrorism. This resignation highlights ongoing tensions within AI safety teams regarding the prioritization of safety over rapid technological progress.

Slashdot Other other

Apple fixes zero-day flaw used in 'extremely sophisticated' attacks

Apple has patched a zero-day vulnerability (CVE-2026-20700) in dyld, affecting multiple devices and exploited in sophisticated attacks. Users are urged to update their devices to prevent potential arbitrary code execution, highlighting the importance of timely security updates in software development.

Bleeping Computer Dev Tooling & DX security apple

Domain Intelligence Wins: What “High-Quality” Actually Means in Production AI

High-quality agentic AI relies on system reliability rather than model sophistication. Organizations should focus on unified data foundations and clear accountability to enhance trust and effectiveness in production AI applications, particularly in regulated environments.

Databricks Blog Building AI Products data-strategy

google/langextract: A Python library for extracting structured information from unstructured text us...

LangExtract is a Python library designed for extracting structured information from unstructured text using LLMs. It features precise source grounding, reliable structured outputs, and optimized strategies for processing long documents, making it adaptable for various domains without requiring model fine-tuning.

GitHub Trending Open Source Drops gemini gemini-ai gemini-api

I Built an AI Game Dev Toolkit (and Cut My Asset Creation Time by 70%)

Sorceress is an AI game development toolkit that streamlines asset creation, reducing time spent on tasks by 70%. Key features include a sprite automator, cost-effective AI art generation, and a free coding agent, all designed to enhance productivity for indie game developers.

Dev.to AI-Assisted Coding ai gamedev productivity

iOfficeAI/AionUi: Free, local, open-source 24/7 Cowork and OpenClaw for Gemini CLI, Claude Code, C...

AionUi is a free, open-source platform that provides a unified graphical interface for various command-line AI tools, enhancing usability and local data security. It supports multi-agent collaboration, real-time file previews, and flexible model switching, making it ideal for AI office automation workflows.

GitHub Trending Open Source Drops acp ai ai-agent

Layoffs! Fellow employees, be prepared for the worst.

GitHub Copilot CLI significantly accelerated the development of a job application management tool by generating project structures, implementing authentication flows, and debugging issues. This experience highlights how AI can enhance coding efficiency and improve code quality while facilitating learning.

Dev.to AI-Assisted Coding devchallenge githubchallenge cli

Microsoft warns that poisoned AI buttons and links may betray your trust

Microsoft warns of AI Recommendation Poisoning, a technique where hidden prompts manipulate AI outputs, leading to biased recommendations. This poses risks in critical areas like health and finance, eroding user trust in AI systems as compromised models may provide misleading information without users' awareness.

The Register Other other

Show HN: Double blind entropy using Drand for verifiably fair randomness

Double blind entropy using Drand offers a method for generating verifiably fair randomness in applications like gaming and lotteries. This approach enhances trust and transparency by combining player and server inputs with a public randomness beacon, ensuring the integrity of random outcomes.

Hacker News Building AI Products building-ai-products

Spotlight on SIG Architecture: API Governance

API Governance in Kubernetes ensures stability while allowing innovation across various API surfaces, including command-line flags and configuration files. The project emphasizes early involvement in design and implementation to maintain conceptual integrity and user trust, adapting guidelines as new scenarios arise.

Kubernetes Blog Dev Tooling & DX dev-tooling-dx

The Collaboration Advantage: How Modern Retailers and CPG Brands Win Together

Delta Sharing revolutionizes data collaboration in retail and CPG by enabling real-time access to live data across platforms without the burdens of traditional ETL processes. This approach enhances decision-making speed, reduces stockouts, and improves promotional ROI, driving significant revenue gains.

Databricks Blog Dev Tooling & DX industries

AI inference startup Modal Labs in talks to raise at $2.5B valuation, sources say

Modal Labs, an AI inference startup, is in talks to raise funds at a $2.5 billion valuation, significantly increasing from its previous $1.1 billion. The company focuses on optimizing inference processes, which enhance efficiency and reduce costs, attracting considerable investor interest in the AI infrastructure space.

TechCrunch Building AI Products enterprise startups ai

Anthropic’s Claude Cowork finally lands on Windows — and it wants to automate your workday

Anthropic's Claude Cowork AI agent is now available on Windows, enhancing file management and task automation capabilities for enterprise users. The integration of Claude's advanced models into Microsoft's ecosystem signals a significant shift in AI tool adoption and competition within enterprise software.

VentureBeat Building AI Products technology orchestration

Building What’s Next. Together. Introducing the Brickbuilder Partner Network for the Agentic AI Era

The Brickbuilder Partner Network aims to enhance collaboration and growth among Databricks partners by introducing a unified tiering structure and a consumption-centric reward model. This initiative focuses on leveraging AI-driven tools and specialized expertise to drive business outcomes in the evolving agentic AI landscape.

Databricks Blog Dev Tooling & DX platform

Coinbase rolls out Agentic Wallets, a wallet infrastructure that lets AI agents make autonomous crypto transactions /'without human intervention/' (Daniel Kuhn/The Block)

Coinbase's Agentic Wallets enable AI agents to autonomously conduct crypto transactions without human oversight, marking a significant shift in wallet infrastructure. This innovation raises important questions about security, trust, and the future of automated financial transactions in the crypto space.

Techmeme Other other

Crazy ransomware gang abuses employee monitoring tool in attacks

A ransomware gang is exploiting legitimate employee monitoring tools to maintain persistence and evade detection in corporate networks. Researchers highlight the use of these tools for remote access and monitoring, emphasizing the need for organizations to enforce MFA and monitor unauthorized installations.

Bleeping Computer Other security

Devilish devs spawn 287 Chrome extensions to flog your browser history to data brokers

A security researcher has identified 287 Chrome extensions that leak browsing history to over 30 data brokers, affecting approximately 37 million users. The findings underscore the urgent need for better user awareness and safeguards against malicious extensions that compromise privacy.

The Register Other other

GitHub availability report: January 2026

GitHub's January 2026 availability report details two significant service outages affecting GitHub Copilot and other core functionalities. The incidents were caused by configuration errors and infrastructure updates, prompting improvements in monitoring and validation processes to enhance future service reliability.

GitHub Blog Dev Tooling & DX company-news news-insights github-availability-report

GLM-5: From Vibe Coding to Agentic Engineering

GLM-5 is a new 754B parameter model that significantly advances the capabilities of LLMs, introducing the concept of 'Agentic Engineering' for professional software engineers. This shift emphasizes the role of engineers in leveraging LLMs for complex coding tasks, showcasing practical applications and improvements over previous models.

Simon Willison AI-Assisted Coding ai-assisted-coding

Is a secure AI assistant possible?

OpenClaw, an AI personal assistant, raises significant security concerns as it requires extensive access to user data, making it vulnerable to prompt injection attacks. Experts emphasize the need for robust security measures to protect users while balancing utility and safety in AI agent deployment.

MIT Technology Review AI Agents & Tool Use artificial-intelligence ai app

Lawmakers demand great wall to keep advanced chipmaking gear out of China

A bipartisan group of US lawmakers is advocating for a comprehensive ban on the sale of advanced semiconductor manufacturing equipment to China, arguing that existing entity-specific restrictions are ineffective. They emphasize the urgency of aligning with allies to prevent China from developing its own chipmaking capabilities.

The Register Other other

Microsoft fixed a vulnerability in Notepad in Windows 11 allowing attackers to execute remote programs by tricking users into clicking malicious Markdown links (Lawrence Abrams/BleepingComputer)

Microsoft addressed a critical vulnerability in Notepad on Windows 11 that could allow attackers to execute remote programs via malicious Markdown links. This fix is crucial for enhancing user security and preventing potential exploitation through social engineering tactics.

Techmeme Other other

Microsoft just forked Windows

Microsoft has forked Windows 11 into two distinct development tracks for Arm and x86 architectures, introducing complexity for consumers and developers alike. This unprecedented move raises questions about future updates and the potential for a new Windows 12, as each track will receive its own patches and updates.

PCWorld Other windows

New Spyware Can Track Everything You Do On Both Android And iPhone - Here's How To Stay Safe

ZeroDayRAT is a new spyware that compromises both Android and iOS devices, providing attackers with extensive access to personal data, including live camera feeds and keylogging capabilities. This malware poses significant risks, especially with its ability to target banking and cryptocurrency applications.

BGR Other news apps-software security

North Korea's UNC1069 Hammers Crypto Firms With AI

North Korea's UNC1069 group is targeting crypto firms by utilizing AI technologies such as large language models and deepfakes, alongside legitimate platforms. This shift highlights the evolving threat landscape in the Web3 space, emphasizing the need for enhanced security measures against sophisticated cyber threats.

Dark Reading Other dev-tooling-dx

Once-hobbled Lumma Stealer is back with lures that are hard to resist

Lumma Stealer has resurfaced with advanced social engineering tactics, specifically using 'ClickFix' to exploit user trust and install malware. This resurgence highlights the importance of user awareness and security measures against evolving infostealer threats targeting sensitive data on Windows systems.

Ars Technica Other biz-it security castleloader

OpenAI Researcher Quits, Warns Its Unprecedented ‘Archive of Human Candor’ Is Dangerous

Zoë Hitzig's resignation from OpenAI highlights concerns over the potential misuse of sensitive user data in advertising on ChatGPT. She warns that the economic incentives may lead to privacy violations, urging for independent oversight to protect user interests amidst growing public apathy towards data privacy.

Gizmodo Other artificial-intelligence

Police arrest seller of JokerOTP MFA passcode capturing tool

The arrest of a seller involved in the JokerOTP phishing automation tool highlights the ongoing threat of MFA bypass techniques used by cybercriminals. This case underscores the importance of user awareness regarding phishing scams and the need for robust security measures to protect sensitive information.

Bleeping Computer Other security

Reports of Telnet's death have been greatly exaggerated

Recent analysis reveals that claims of widespread Telnet traffic blocking by core ISPs are unfounded. Data from multiple sources, including RIPE Atlas and internal measurements, indicate continued Telnet activity, suggesting that observed drops may stem from specific scanning behaviors rather than ISP-level censorship.

Hacker News (Official) Other other

Show HN: Agent framework that generates its own topology and evolves at runtime

The agent framework generates its own topology and evolves at runtime, enabling dynamic adaptation to changing environments. This innovative approach enhances the capabilities of AI agents in real-time applications, making them more efficient and responsive to user needs.

Hacker News (Official) AI Agents & Tool Use other

Skills in OpenAI API

OpenAI's Skills feature allows users to integrate custom skills directly into the API using inline base64-encoded zip data. This capability enhances the flexibility of the API, enabling developers to create more dynamic and interactive applications with ease.

Simon Willison AI-Assisted Coding ai-assisted-coding

T-Mobile’s new network-level AI translates calls on every phone

T-Mobile has integrated a real-time AI translation feature into its network, enabling live translations during calls across 50+ languages without requiring new devices or apps. This innovation aims to enhance accessibility and communication, potentially setting a precedent for other carriers to follow.

Android Police Building AI Products t-mobile carriers

The Policy Risk Of Closing Off New Paths To Value Too Early

Artificial intelligence is reshaping societal perceptions of work value, necessitating a reevaluation of how new forms of labor are legitimized. Rapid technological change demands that policymakers create space for experimentation with new economic activities to adapt effectively without prematurely closing off potential value pathways.

Techdirt Other 1 ai business-models

The Problem with LLMs

LLMs raise ethical concerns around plagiarism and copyright violations, complicating their integration into software development. While they enhance accessibility and streamline workflows, the risk of introducing flawed code and the phenomenon of 'AI Fatigue' highlight the need for cautious adoption and ongoing reflection on their impact.

Hacker News Building AI Products building-ai-products

Two more xAI co-founders are among those leaving after the SpaceX merger

Following the recent SpaceX merger, two co-founders of xAI have departed, leaving the company with half of its original leadership. Elon Musk announced a reorganization aimed at enhancing execution speed, introducing four divisions focused on AI applications, including space-based initiatives.

The Verge Other other

xAI lays out interplanetary ambitions in public all-hands

xAI's recent all-hands meeting revealed ambitious plans for AI product development, including a new organizational structure and significant usage metrics. Elon Musk emphasized the potential for space-based data centers and AI satellite manufacturing, showcasing a forward-thinking vision for AI's future.

TechCrunch Building AI Products ai elon-musk xai

Heroku is not dead

Heroku's challenges post-acquisition by Salesforce highlight the complexities of organizational dynamics and resource management in tech companies. The insights shared reflect on the inefficiencies and systemic issues that arise when scaling operations, emphasizing the need for data-driven decision-making and effective integration strategies.

Hacker News Building AI Products building-ai-products

From 34% to 96%: The Porting Initiative Delivers – Hologram v0.7.0

Hologram v0.7.0 significantly enhances browser-side Elixir capabilities, increasing Erlang function coverage from 34% to 96%. This release includes over 150 newly ported functions, improving string processing, collections, and math operations, while also addressing performance and compatibility enhancements.

Hacker News Building AI Products building-ai-products

3 takeaways from Elon Musk's xAI all-hands, from a moon city to a company restructuring

Elon Musk's xAI all-hands meeting highlighted ambitious plans for a moon city and significant company restructuring. These developments indicate a shift in focus for xAI, potentially impacting its future projects and collaborations in the AI landscape.

Business Insider Other other

DIY Wall-Plotter Does Generative Art, But Not As We Know It

GPenT is a wall-mounted polargraph that generates line art using various methods, including a diffusion model that converts text directly into G-code. This project combines machine learning with CNC technology, showcasing unique generative art capabilities and a comprehensive software pipeline.

Hackaday Other art machine-learning generative-art

Anthropic says it’ll try to keep its data centers from raising electricity costs

Anthropic pledges to mitigate the impact of its data centers on local electricity costs by covering infrastructure upgrades and committing to reduce power consumption during peak demand. This initiative reflects a growing trend among tech companies to address the environmental and economic concerns associated with energy-intensive AI operations.

The Verge Other other

Applied Materials reaches a $252M settlement with the US Commerce Department, resolving allegations it illegally shipped chipmaking equipment to SMIC in China (Reuters)

Applied Materials has settled with the US Commerce Department for $252M over allegations of illegally shipping chipmaking equipment to SMIC in China. This settlement highlights the ongoing tensions in the semiconductor industry and the regulatory scrutiny surrounding technology exports.

Techmeme Other other

Is Linux Mint Burning Out? Developers Consider Longer Release Cycle

Linux Mint developers are considering a longer release cycle to allow for deeper development work, moving away from their current six-month cadence. This potential shift aims to free up resources for more substantial improvements, reflecting on the challenges faced by a small team in maintaining consistent releases.

Slashdot Other other

Quoting Andrew Deck for Niemen Lab

The New York Times developed an AI tool called the 'Manosphere Report' that utilizes large language models to summarize and transcribe podcasts, providing timely insights into conservative media sentiment. This innovative application of AI enhances journalistic coverage by delivering rapid analysis of emerging narratives.

Simon Willison Building AI Products ai-assisted-coding

SpaceX Is Becoming The Most Valuable Private Company In The World With New Merger

SpaceX's merger with xAI positions it as the most valuable private company, valued at around $1 trillion. This strategic move aims to address the growing power demands of AI data centers by exploring space-based solutions, potentially revolutionizing AI infrastructure.

BGR Other space

Uber Eats launches AI assistant to help with grocery cart creation

Uber Eats has launched an AI-powered Cart Assistant to streamline grocery shopping by allowing users to upload lists or images for automatic cart creation. This feature enhances personalization by prioritizing familiar items based on previous orders, positioning Uber Eats competitively against other delivery services integrating AI.

TechCrunch Building AI Products ai apps ai-chatbot

What Happens When a Smart Hub Processes AI Locally Instead of in the Cloud

SwitchBot's AI Hub processes video feeds locally, eliminating the need for cloud infrastructure and enhancing privacy. It supports real-time object recognition and automations through a user-friendly interface, appealing to those seeking advanced smart home capabilities without ongoing cloud costs.

The Gadgeteer Local & Edge AI news camera security-camera