Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Laguna XS.2 and Laguna M.1 are agentic coding models built for long-horizon work. Laguna M.1 is the foundation for the Laguna family of models. Laguna XS.2 is a much smaller model but is still remarkably capable for its size. Both models are free to use for a limited time via Poolside's API and on OpenRouter. The weights for Laguna XS.2 have been released under an Apache 2.0 license.
The Recurrent Transformer is a novel architecture designed to overcome the "temporal shallowness" of standard Transformers. By introducing layerwise recurrence, the model uses a combination of temporary and persistent key-value pairs to allow each layer to maintain a continuously updating internal memory. This mechanism enables the model to perform complex, iterative reasoning across a sequence without requiring massive layer stacks.
Shares of several firms tied to OpenAI dropped on Tuesday after news that OpenAI had missed its own targets for revenue and users surfaced. Investors are worried that OpenAI's massive investment in AI may not produce the blockbuster profits many expect. OpenAI critics have said that certain financing arrangements the company has made are circular in nature, with the company's partners providing funding and the company spending money on computing with that partner. The company has defended its financial footing and says its leaders are aligned on securing computing resources.
Meta's new AI model, Muse Spark, signals a shift in its AI strategy, moving from open-source to paid access, aiming to boost its ad business and compete in the growing AI market. Analysts are cautiously optimistic, noting Meta's internal testing and investments, despite trailing top performers like Anthropic's Claude. Meta's workforce reduction and strategic hires reflect its intensified focus on AI advancements and infrastructure.
Sponsored
As agentic systems grow, coordination, integrations, testing, and data handling become architectural problems. Explore how neuro-san is designed to handle these constraints so multi-agent systems remain scalable and reliable. Read post [19].
Parallel Web Systems, a startup founded by Parag Agrawal, the former chief executive of Twitter, has developed a platform for AI agents to search the web.
Agent Templates on the ElevenAgents platform enables quick deployment of AI agents with pre-built frameworks.
TLDR is looking for an engineer/researcher at a major AI lab or startup to help write for 1M+ subscribers. Our curators have been invited to Google I/O and OpenAI DevDay, scouted for Tier 1 VCs, and get early access to unreleased TLDR products. Learn more [23].
Google agreed to provide the US Department of Defense access to its AI on classified networks for broad lawful uses.
Meta Sapiens2 is a family of high-resolution transformers trained on 1B human images that achieves strong results across tasks like pose estimation, segmentation, and surface prediction.
Snapchat is rolling out “AI Sponsored Snaps,” a new ad format that lets users chat directly with brands' AI agents inside the app's Chat tab, allowing them to ask questions, get product recommendations, and interact with ads conversationally. The company says the feature builds on the success of Sponsored Snaps, which already drive higher conversions at lower costs, and reflects Snap's belief that chat-based interactions are becoming a major part of digital advertising as users increasingly engage with messaging and AI-powered conversations inside the app.
Under new leadership, Xbox is redefining its strategy around being "affordable, personal, and open," targeting daily active players as its north star and expanding across console, PC, mobile, and cloud. CEO Asha Sharma and Xbox Game Studios head Matt Booty have acknowledged failures in areas like pricing, PC presence, and core platform features, while flagging rising development costs and indie competition as growing challenges. Exclusivity, windowing, and AI are also under review. The Microsoft Gaming brand is being retired in favor of the Xbox name.
Lovable has launched a vibe-coding app for iOS and Android that allows users to build websites and web apps through AI-powered voice or text prompts while moving seamlessly between mobile and desktop workflows. The launch follows Apple's recent crackdown on vibe-coding apps like Replit and Vibecode, which were forced to stop running generated apps directly inside their mobile apps due to App Store security rules. To comply, Lovable appears to have shifted previews to the web instead.
Apple's UX strategy isn't just aesthetic minimalism — it's a behavioral system built around five interlocking principles (usability, communication, functionality, aesthetics, and emotional design) that together eliminate friction and guide users toward confident decisions. Product pages function as structured narratives, distributing cognitive effort progressively so that, by the time users reach a call to action, purchasing feels like a natural conclusion rather than a deliberate choice. This approach is credited with driving conversion rates of 5–10%, well above the typical e-commerce average of 2–3%.
AI design tools that ignore your design system create visual drift, component debt, and governance erosion by generating designs that look similar to your components but aren't built with your actual design system. The cleanup required to align AI-generated designs with existing design systems often takes longer than building from scratch. Most AI tools lack connection to component libraries, generating output that appears correct but bypasses established design constraints and contribution processes.
Design isn't about producing outputs but about deeply understanding a problem and achieving a good fit between form and context. Tools (including AI) can generate results quickly, but they don't replace the thinking required to resolve underlying complexities. Overreliance on AI risks creating polished yet fragile products, because real design value comes from the slow, iterative process that clarifies understanding—not just the final output.
TUIStudio is a Figma-like visual editor for designing terminal user interfaces (TUIs) with drag-and-drop components and real-time ANSI preview.
Transform your imagination into studio-grade visuals. Input any image and see it refracted through Doodles lens in seconds.
A Claude-powered Figma plugin built around an agent loop — not a chat window.
AI can analyze usability test videos by sampling frames, transcribing audio, and generating plausible descriptions of user behavior — though this process is "lossy" and prone to inconsistency. In a test using a six-minute OpenTable usability session, ChatGPT identified seven specific problems, including a search field that dropped cuisine selections when the location was changed and results that surfaced non-sushi restaurants. Whether those findings hold up in terms of accuracy, reliability, and alignment with what human researchers would catch remains an open question to be addressed in follow-up research.
Apple's launch of the low-cost MacBook Neo has been highly successful. A behind-the-scenes video revealed that its polished marketing relied heavily on practical, handcrafted effects combined with some CGI. Real-world techniques and artists were central to the production.
The Mosaic Rooms' new identity by Samar Maakaroun intentionally rejects stability, using a fluid monogram and bold color choices to express movement, cultural duality, and the idea of being “in-between” rather than fixed. The rebrand reflects the institution's evolving role and shows how strong design can embody unresolved tension and complex identity instead of simplifying it.
PITAKA is launching a global design competition, "Weave the Next, Weave Our World," from April 24 to May 25.
A well-designed favicon increases credibility, improves user experience by helping visitors navigate multiple tabs, and maintains brand consistency across digital platforms.
IKEA's “Unpackaged Goods” campaign uses minimalist, low-noise visuals to highlight the satisfaction of organization.
Sponsored
You've been there: code's ready to ship, everyone's excited—then someone spots a vulnerability. Suddenly it's an all-hands emergency and time to cancel your evening plans.
The new CSS `animation-timeline` API allows devs to create scroll-driven animations natively in CSS, moving away from JavaScript solutions. The API extends existing CSS keyframe animations by mapping their progress to an element's visibility within the viewport or the document's overall scroll position. These animations can be finely controlled using various timing functions, precise animation ranges like `entry` or `exit`, and even link animations across different elements.
Mendral reduced costs by using a multi-agent architecture where a cheap "Haiku triager" handles 80% of routine CI log failures, escalating only complex issues to a "frontier" Opus model. This tiered system allows the expensive model to focus on orchestration and delegation while low-cost agents perform the data-heavy tasks, resulting in significantly more efficient LLM operations.
Too many MCP tools bloat context and cause hallucinations. Two fixes include reducing tools (filter agent-side or design fewer use-case-driven ones MCP-side), or use “code mode.” This means give the LLM the ability to search and execute, let it write code against a generated SDK, and run multi-step workflows in a sandbox.
Mitchell Hashimoto is moving his project, Ghostty, off GitHub after 18 years due to frequent outages that hinder professional collaboration. Although deeply attached to the platform, he believes its current lack of reliability makes it unsuitable for serious work until significant improvements are made.
GitHub Actions has become a major vulnerability in the open-source supply chain due to insecure defaults that malicious actors exploit to inject malware and steal credentials. While GitHub proposes opt-in fixes, maintainers must currently rely on third-party tools and rigorous configuration to mitigate these significant security risks.
Sponsored
Writing code can be stressful—but not half as stressful as a surprise security meltdown. Inject optimism and calm into the developer scrum with Microsoft Azure [11]. Unified security across code and cloud environments and built-in DDoS protection mean you've got less cause for concern—and a clear mind for innovation. Help
Warp has open-sourced its client to pioneer an agent-first workflow powered by its Oz platform and GPT models.
VibeVoice is an open-source frontier voice AI project by Microsoft that provides models for Text-to-Speech and Automatic Speech Recognition, using continuous speech tokenizers and a next-token diffusion framework with LLMs for high-fidelity processing of long audio. The project features VibeVoice-ASR for multilingual recognition and VibeVoice-TTS for multi-speaker and streaming synthesis.
Legal ownership of AI-generated code is complex, as it may not qualify for copyright without meaningful human authorship and could inadvertently include open-source licensed material. To manage these risks, developers must carefully document their contributions, review employment contracts, and use licensing tools to ensure compliance and protect intellectual property.
OpenAI and AWS have formed an exclusive partnership to launch Bedrock Managed Agents, integrating frontier models with AWS infrastructure to simplify the deployment of virtual co-worker agents for enterprise clients. Both CEOs view this as a new paradigm in agentic computing that will reduce complexity and cost while keeping customer data secure within AWS.
This dev scraped all 84,000 Firefox extensions and tried to install them simultaneously, surfacing fun finds along the way (a 196 MB chess extension, one requesting 3,695 permissions, and a PUA network with 700K+ users hijacking search to a Yahoo affiliate code). After ten failed attempts, a friend's beefier VM finally booted Firefox with all of them. Extensions.json hit 189 MB, memory sat at 27–37 GB, about:addons took 6 hours to load, and example.com never loaded.
Anthropic is launching new connectors to integrate Claude with popular creative software like Adobe and Blender, empowering professionals to streamline workflows, accelerate ideation, and automate repetitive tasks.
ChatGPT serves contextual ads by injecting them into conversations and tracks user engagement on merchant websites using a proprietary SDK and encrypted attribution tokens.
GitHub's current instability necessitates a publicly funded, independent archive to preserve open-source software history beyond a single platform.
Structured Prompt-Driven Development (SPDD) is a method that treats AI prompts as first-class, version-controlled artifacts to make sure LLM-assisted software changes are governable, reviewable, and reusable for team-based delivery.
Sponsored
Many founders don't. But with HighLevel [4], you can stop juggling 10+ marketing, sales, & automations solutions and get things done again.
"Does it integrate with Salesforce?" was the question buyers asked for a decade. Now, they want to know if agents can drive your product, if your APIs are clean, and if there's an MCP connector. Buyers asking those questions are buyers running the math on whether to keep you. The math gets worse when a vendor they already pay adds a 60% version of what you do, because AI fills enough of the rest that cutting your $80K contract stops feeling like a tradeoff. Annual deals have hidden how often this is already happening.
AI agents now execute the actual work as code loops rather than just mediating interfaces, converting human jobs into tool calls and verification steps. Agentic tasks burn orders of magnitude more tokens than chat because a single Claude Code bugfix can consume around 900k tokens, where almost all goes to context replay and tool output rather than visible code. METR data shows autonomous task horizons doubling every 131 days, jumping from 4 minutes on GPT-4 to roughly 12 hours on Claude Opus 4.6 and pushing OpenAI past 15B tokens per minute alongside Google's 50x YoY token growth.
Chainguard now expects its engineering managers to be at the 50th percentile of token usage among their direct reports. Managers who use too few tokens lack the context needed to manage and move results forward. Managers who use too many tokens need to focus on increasing token usage amongst their teams. Right around the middle means that both managers and their teams understand what their tools can do.
Polished prose used to take effort. Now, any founder can produce a clean feature announcement in ten minutes, but most of them read like the same announcement. Swap your company name for a competitor's and read your last post again. If it still makes sense, you wrote a press release. Don't hand the model a feature spec and ask for a polished post. Hand it the customer ticket, the Slack argument, and the voicemail where you explain why this even matters.
Sponsored
Microsoft Build, June 2-3 in San Francisco [10], is designed for the product-minded developer and technical entrepreneur that wants to learn from the teams at the forefront of AI. Through hands-on demos and labs get the technical depth you need to create what's next. Spots