Tag: Meta

  • Meta Introduces AI-Powered Alerts to Protect Teens from Self-Harm Conversations

    Meta Introduces AI-Powered Alerts to Protect Teens from Self-Harm Conversations

    Meta is rolling out a new safety feature that alerts parents when their supervised teen discusses suicide or self-harm with Meta AI. The company has also updated how its AI responds to such conversations and expanded content restrictions for teen accounts.

    The feature is currently launching in the United States, the United Kingdom, Australia, and Canada, with plans for a global rollout by the end of the year.

    Parents to Receive Safety Alerts

    If a supervised teen shares thoughts of suicide or self-harm, Meta AI will encourage them to reach out to a trusted adult, counselor, or crisis support service. Parents using Meta’s supervision tools will receive a notification if their teen needs support. The company aims to help parents recognize signs of distress and intervene when necessary. The system can identify conversations where a teen may be expressing suicidal thoughts, even if the language is not explicit.

    Every Alert Will Be Reviewed

    Meta emphasized that alerts are not generated automatically. Each conversation flagged by its AI is reviewed before a notification is sent to a parent. The company said it will ‘err on the side of caution’ when assessing conversations. If there is uncertainty, parents may still receive an alert rather than risk missing a genuine case where a teen needs help.

    AI Responses Updated with Expert Input

    To improve Meta AI’s responses, the company consulted more than 75 mental health clinicians with experience in adolescent care. Experts reviewed hundreds of conversations involving teens and evaluated whether the AI responded appropriately. Their feedback was used to enhance how Meta AI handles sensitive discussions.

    Meta said the updated system is designed to acknowledge a teen’s emotions, encourage them to seek help, and connect them with relevant support instead of ending the conversation quickly.

    Emergency Response Feature in Development

    Meta is also developing a feature that could allow it to contact emergency services if conversations with Meta AI suggest a user faces an imminent risk of suicide. The company already reports credible suicide threats identified on Facebook and Instagram. Meta said it made more than 19,000 referrals to emergency responders worldwide last year, enabling authorities to perform welfare checks when required.

    More Restrictions for Teen Accounts

    Meta is extending Instagram’s Limited Content setting to Meta AI chats. Teen Accounts already receive age-appropriate AI responses. Meta AI does not engage in sexual or romantic conversations with teens and does not provide recipes for alcoholic drinks. With the stricter Limited Content setting enabled, Meta AI will refuse a wider range of sensitive requests.

    The update builds on Meta’s existing parental supervision tools, including alerts when supervised teens repeatedly search for suicide or self-harm-related content on Instagram. The company said the latest changes are aimed at giving parents more visibility while strengthening protections for teenagers using its AI services.

  • WhatsApp Communities Update Hides Numbers from Strangers but Not Admins

    WhatsApp Communities Update Hides Numbers from Strangers but Not Admins

    WhatsApp’s latest update for Communities introduces a privacy feature that allows users to hide their phone numbers from strangers in large groups. However, the feature does not offer complete anonymity. Community admins, saved contacts, and anyone with a previous chat history can still view members’ phone numbers.

    Meta launched this privacy update in July 2023, aiming to protect users during large community discussions. The feature works exclusively within Communities, which can support thousands of members. In standard WhatsApp groups, all participants can still see each other’s phone numbers unless separate privacy settings are applied.

    Meta designed the update to reduce unnecessary exposure while maintaining moderation tools for administrators. Every community admin retains full access to member phone numbers, enabling them to moderate, manage, and communicate as needed. Saved contacts and individuals who have had direct conversations with a member also continue to see phone numbers without restrictions.

    The privacy feature has important limitations that users should consider before joining a Community. If a user joins a standard WhatsApp group, these protections are removed. Additionally, Community privacy never hides numbers from administrators or existing contacts. Members cannot choose specific people who should or should not view their numbers; the system operates under fixed visibility rules rather than custom controls.

    Privacy experts have raised concerns as WhatsApp expands identity protection features. Reporting indicates that username-based messaging, currently under development, could further reduce phone number sharing. However, Indian regulators have expressed concerns about potential impersonation and online fraud risks before a wider rollout.

    WhatsApp’s Community-based privacy rules offer stronger protection than traditional group chats, but the feature should not be viewed as complete anonymity. Members still share phone numbers with admins, trusted contacts, and previous chat connections. This update represents a practical privacy improvement rather than a full identity shield.

  • WhatsApp Tests Independent Cloud Backup for iPhone Users, Reducing iCloud Reliance

    WhatsApp Tests Independent Cloud Backup for iPhone Users, Reducing iCloud Reliance

    WhatsApp is developing a new feature that would allow iPhone users to back up their chat history without relying on Apple’s iCloud. The move, discovered in the latest WhatsApp beta for iOS by WABetaInfo, points to Meta’s plan to introduce its own encrypted cloud storage service for conversation backups.

    Currently, iPhone users must use iCloud for WhatsApp backups, requiring sufficient iCloud storage space. The upcoming change would give users the option to choose between iCloud and WhatsApp’s own cloud service, potentially easing storage constraints for those who frequently run out of space or prefer not to depend on Apple’s ecosystem.

    According to leaked screenshots, the new system will offer 2GB of free cloud storage from WhatsApp, with additional paid plans likely available for users who need more capacity. Like the existing backup option, WhatsApp’s cloud backups will be end-to-end encrypted, ensuring messages remain private and secure.

    The feature is still in development and not yet accessible to beta testers. Meta has not announced a release date or availability timeline. If launched, it would give iPhone users greater flexibility in managing their chat backups and reduce dependence on iCloud for storage.

  • Meta Denies Using AI to Decide Layoffs Amid Landmark Lawsuit

    Meta Denies Using AI to Decide Layoffs Amid Landmark Lawsuit

    Meta has officially denied allegations that it used artificial intelligence to determine which employees would be laid off during its May 2026 restructuring, which eliminated about 8,000 jobs—roughly 10% of its global workforce. The company insists that human managers made all workforce decisions, and that AI only supported administrative processes.

    The denial comes in response to a lawsuit filed on July 14 in the US District Court for the Northern District of California by 26 current and former employees. The plaintiffs accuse Meta of using an internal AI system called Metamate—a large language model trained on employee communications and documents—alongside algorithmic productivity scores derived from keystroke activity, browser history, email data, and AI token consumption. They claim this system scored, ranked, and selected workers for termination in a way that disproportionately targeted people on medical, maternity, or parental leave.

    Nearly half of the plaintiffs say their termination directly followed protected leave. The complaint argues that the scoring framework failed to account for time spent on legally protected leave, effectively penalizing employees for reduced output during periods when they were legally entitled to be away. “Meta did not assemble the termination list through the considered judgment of managers who knew the work,” the complaint states.

    Meta’s spokesperson told CNBC that the claims “lack merit and are not based on facts,” maintaining that human judgment drove all workforce decisions. The plaintiffs are now seeking a preliminary court ruling to block Meta from completing the layoffs on July 22 while they pursue their claims in private arbitration, as required by their employment contracts.

    Legal analysts note that a human manager signing off on an algorithm’s output does not automatically insulate a company from discrimination claims if the underlying scoring criteria were discriminatory. This lawsuit arrives just weeks after a federal judge in California ruled that Workday must face similar claims that its AI-powered job-screening services allegedly violate anti-discrimination laws.

    This case is the first of its kind against a major US company regarding AI-selected layoffs. Its outcome could shape how courts evaluate the use of AI in workforce decisions, what disclosure obligations employers carry when algorithmic tools inform layoffs, and whether “a human approved it” constitutes sufficient oversight when the scoring methodology was built without legal-leave exemptions. For the broader tech industry, which has rapidly adopted AI-driven performance tools, the case represents an early signal of legal and regulatory exposure.

  • Leading Smart Glasses Brands Shaping 2026: Meta, XREAL, Apple & More

    Leading Smart Glasses Brands Shaping 2026: Meta, XREAL, Apple & More

    The smart glasses market is rapidly evolving, with 2026 poised to be a breakthrough year for wearable AI and augmented reality. From stylish consumer frames to enterprise-grade headsets, several brands are leading the charge. Here’s a look at the companies defining the future of smart eyewear.

    Meta Ray-Ban

    Meta’s collaboration with Ray-Ban has produced stylish smart glasses integrated with AI features, a camera, speakers, and voice assistance. These glasses blend fashion with functionality, making them a popular choice for everyday use.

    XREAL

    XREAL (formerly known as Nreal) offers popular AR glasses designed for immersive entertainment and productivity. Their lightweight form factor and compatibility with smartphones make them a leader in consumer AR.

    Google

    Google continues to develop AI-powered smart eyewear that integrates with its Gemini AI platform. Leveraging years of research, Google aims to bring contextual assistance and hands-free computing to users.

    Apple

    Apple’s wearable roadmap is expected to redefine premium mixed-reality experiences. While the Apple Vision Pro sets the stage, future iterations may bring more accessible smart glasses, combining high-end design with advanced sensors.

    Vuzix

    Vuzix focuses on enterprise-grade smart glasses for logistics, healthcare, and industrial applications. Their headsets provide rugged, hands-free solutions for workers requiring real-time data overlays.

    Lenovo

    Lenovo offers innovative AR wearables designed for business collaboration and virtual workspaces. Their ThinkReality series targets professionals seeking augmented productivity tools.

    Rokid

    Rokid produces affordable AR smart glasses that are gaining popularity for education and entertainment. Their devices focus on accessibility, making AR experiences available to a wider audience.

    Samsung

    Samsung is building an AI-powered XR ecosystem expected to expand wearable computing. With deep expertise in displays and mobile technology, Samsung’s upcoming smart glasses could integrate seamlessly with Galaxy devices.

    These brands are pushing the boundaries of what wearable AI and AR can achieve, from personal assistants to immersive work tools. Keep an eye on 2026 as these companies release next-generation smart glasses.

  • Master Dynamic Programming in 2026: Top Patterns and Interview Questions

    Master Dynamic Programming in 2026: Top Patterns and Interview Questions

    Dynamic programming remains one of the most important coding interview topics in 2026. Mastering common DP patterns, understanding company-specific expectations, and practicing representative problems can significantly improve problem-solving skills and increase success in technical interviews at top technology companies.

    Why Pattern Recognition Beats Memorization

    Interview data from recent hiring cycles suggests that around 87% of coding questions at Google, Meta, Apple, Netflix, and Amazon are built around just 10 to 12 core problem-solving patterns, with dynamic programming forming one of the most versatile among them. The core idea behind DP is simple even when the problems aren’t: break a larger problem into smaller, overlapping subproblems, solve each one exactly once, and store the results, typically through memoization or a tabulation array, so the same subproblem is never recomputed. Any time a question asks for a maximum, minimum, or count of ways to do something under constraints, there’s a strong chance dynamic programming applies.

    The Core Patterns Worth Mastering

    DP breaks down into a manageable number of recognizable categories rather than one giant, undifferentiated topic. Knapsack-style problems (0/1 and unbounded variants) cover resource allocation and optimization questions. Grid-based 2D DP tests a candidate’s ability to reason across two-dimensional state spaces with multiple transition directions. String DP handles text-processing challenges like edit distance and longest common subsequence, patterns with direct application in areas like version control and bioinformatics. Interval DP applies to scheduling and matrix chain multiplication-style problems, while state-machine DP models transitions governed by specific rules, useful in problems involving buy/sell timing or step-based decisions.

    Frequently Asked DP Problems in 2026

    Company-Specific Differences Worth Knowing: Every organization does not follow the same approach toward DP problems. The first difference between companies is that Google requires higher DP expertise compared to other organizations, not only in 1D problems such as staircase problems but 2D Grid, String DP and even DP on Trees/Graphs problems and an overall preference for harder problems in general. The next two organizations, Meta and Amazon, have more focus on array and string DP problems with medium-level difficulty. Microsoft avoids going into hard problems related to DP and prefers medium-level difficult problems along with good explanations. This shows that preparation according to the requirements of a particular company pays off in a bigger way.

    How to Actually Prepare for DP Interviews

    The most optimal form of preparation is no different from the tried and tested methods of pattern-based preparation. The method is to select a canonical example for each pattern, practice solving it until the logic of the solution takes under two minutes to describe and then use that logic for other variations. Interviewers are much more concerned about the ability of candidates to recognize the structure of subproblems, verbally derive a recurrence relation for the same and explain their trade-offs on time and space than knowing whether the candidate has seen the problem or not. Considering the frequent appearance of DP questions on arrays, strings, trees and grids, spending the effort to really comprehend a few basic patterns goes a long way.

    Why This Matters

    Dynamic programming is one of the most frequently tested problem-solving techniques in software engineering interviews. Understanding its recurring patterns helps candidates solve complex coding challenges efficiently, improve algorithmic thinking, and perform better in technical interviews across leading technology companies.

    FAQs

    What is dynamic programming in programming?

    Dynamic programming (DP) is an algorithmic technique used to solve complex problems by breaking them into smaller overlapping subproblems. Instead of solving the same subproblem repeatedly, DP stores previously computed results using memoization or tabulation, significantly improving time efficiency.

    Why is dynamic programming important for coding interviews?

    Dynamic programming tests a candidate’s problem-solving ability, optimization skills, and understanding of algorithms. Many leading technology companies include DP questions because they evaluate logical thinking, recurrence formulation, state transitions, and the ability to optimize brute-force solutions into efficient algorithms.

    Which companies frequently ask dynamic programming questions?

    Dynamic programming questions commonly appear in interviews at Google, Meta, Amazon, Microsoft, Apple, Netflix, Adobe, Uber, and many other technology companies. Google generally asks more advanced DP problems, while Amazon and Meta often focus on medium-difficulty array and string-based DP questions.

    What are the most common dynamic programming patterns?

    The most frequently tested DP patterns include 0/1 Knapsack, Unbounded Knapsack, Grid DP, String DP, Interval DP, State Machine DP, Tree DP, and Bitmask DP. Learning these core patterns enables candidates to solve a wide variety of interview problems efficiently.

    How should beginners start learning dynamic programming?

    Beginners should first understand recursion before learning memoization and tabulation. They should then practice simple problems like Fibonacci, Climbing Stairs, and House Robber before progressing to Knapsack, Longest Common Subsequence, Coin Change, and more advanced DP patterns.

  • Meta Stock Surges on AI Monetization Plans and Data Center Expansion in 2026

    Meta Stock Surges on AI Monetization Plans and Data Center Expansion in 2026

    Meta Platforms stock rebounded strongly on Friday as investors welcomed fresh details about the company’s artificial intelligence revenue strategy, data center investments, and competitive model pricing. Shares climbed approximately 6% on the day and posted a nearly 15% weekly gain—the best weekly performance since early 2024.

    The rally erased Meta’s year-to-date losses. The stock had been under pressure due to concerns over rising capital expenditure and slower-than-expected returns from AI investments. The latest updates provided Wall Street with clearer signals on how Meta intends to convert its heavy infrastructure spending into new revenue streams beyond advertising.

    Meta Explores AI Cloud Business

    CEO Mark Zuckerberg revealed that Meta is considering renting AI computing power to external customers, noting that demand for the company’s compute capacity is “so high that it may make sense” to offer access through a cloud business. While the plan remains in early stages, it could involve providing access to AI chips and servers or hosting models for enterprises that need additional computing power. This move would position Meta in direct competition with Amazon Web Services, Microsoft Azure, and Google Cloud.

    Demand for computing capacity is rising as users adopt chatbots, coding tools, and agentic systems. Meta already invests heavily in data centers to support Facebook, Instagram, advertising, and AI tools. A cloud offering could generate revenue from the same infrastructure that supports its core platforms.

    Muse Spark 1.1 Introduces Paid Model Access

    Meta released Muse Spark 1.1, a new AI model optimized for coding and agentic workloads, and introduced pricing for developer access. The company will charge $1.25 per million input tokens and $4.25 per million output tokens—significantly lower than Anthropic’s Claude Opus 4.8, which costs $5 per million input tokens and $25 per million output tokens. This aggressive pricing may appeal to developers seeking cost-effective tools for coding, workflow automation, and AI agents.

    Tokens are units of text processed by AI models. Input tokens cover user prompts and commands, while output tokens cover the model’s responses. Lower token prices help developers manage costs when applications handle large volumes of text.

    Earlier in the week, Meta also launched Muse Image, an image generation model targeting creators, advertisers, and subscription products. Together, Muse Image and Muse Spark 1.1 demonstrate Meta’s expanding portfolio of AI tools for both business users and developers.

    Data Centers and Custom Chips Shape the Spending Debate

    Meta announced a new data center in Alberta, Canada, as part of its global infrastructure build-out. The facility will support the company’s growing needs for AI training and model delivery.

    Meanwhile, Meta is advancing its custom AI chip, Iris, with production expected to begin in September. Iris is part of Meta’s effort to reduce costs and gain greater control over the hardware powering its AI systems.

    Capital spending remains a key investor concern. Meta previously guided 2026 capital expenditure as high as $145 billion, raising questions about the pace at which AI investments will generate returns. However, Bank of America analyst Justin Post noted that Meta “may have engineered significant cost savings” in data center capacity, supporting the view that the company could manage its AI build-out at a lower cost than anticipated.

    These developments—combined with AI cloud services, competitive model pricing, infrastructure expansion, and custom chip progress—have renewed investor confidence in Meta’s ability to turn its AI strategy into tangible revenue growth.

  • Meta Suspends Instagram AI Tool Amid User Privacy Backlash

    Meta Suspends Instagram AI Tool Amid User Privacy Backlash

    Meta has paused the rollout of a new AI-powered image feature on Instagram just days after its launch, following a wave of privacy complaints from users. The company stated it is reviewing the feedback and plans to make improvements before reintroducing the tool.

    Privacy Concerns Spark Rapid Suspension

    The feature allowed users to generate AI-created images by referencing photos from their Instagram accounts. While some users found it engaging, many demanded clearer explanations about data usage and image handling. Concerns spread quickly across social media, prompting Meta to halt the feature.

    “Our intent was to provide a useful creative tool and to give people control over whether their public content could be referenced in this way,” a Meta spokesperson said. The company has not announced a timeline for the feature’s return.

    AI Rollbacks Becoming Common Industry Pattern

    Meta’s suspension follows similar incidents at other tech giants. Google faced criticism after AI Overviews returned inaccurate search answers, and Microsoft delayed its Recall feature due to privacy and security concerns. These examples illustrate the pressure companies face to release AI tools quickly, sometimes at the cost of user trust.

    Launching AI features prematurely can lead to expensive fixes and reputational damage. Regulators are also taking closer interest in how such products are developed. Slowing down the release process may ultimately save companies from larger problems.

  • S&P 500 Heads for Weekly Gain as Meta Rallies on AI Cost Savings and Circle Soars on OCC Nod

    S&P 500 Heads for Weekly Gain as Meta Rallies on AI Cost Savings and Circle Soars on OCC Nod

    The U.S. stock market moved in a narrow range on Friday, with the S&P 500 holding near the flatline as Wall Street wrapped up a largely positive week. The benchmark index is on track for a weekly gain of nearly 1%, while the Nasdaq Composite is set to rise more than 1%. The Dow Jones Industrial Average lagged slightly, heading for a weekly loss of less than 1%.

    Investors weighed a sharp rally from the previous session, driven by cooling oil prices and a temporary easing of Middle East tensions. President Donald Trump said Iran had called to make a deal, and officials from Qatar and Pakistan are working to restart talks between Washington and Tehran. “The muted stock market reaction to the re-escalation of Iran tensions this week is prime evidence that the market is looking past geopolitical tensions,” said Clark Bellin, president and chief investment officer at Bellwether Wealth.

    Chip Stocks Weaken Ahead of SK Hynix NASDAQ Debut

    Chip stocks traded lower ahead of SK Hynix’s U.S. market debut on the Nasdaq. The South Korean memory-chip maker priced its American depositary receipts at $149 each, raising $26.5 billion. The listing underscores strong investor demand for a major supplier in the AI chip supply chain. Micron Technology fell around 3%, while Marvell Technology and Lam Research lost more than 1%. Intel dropped nearly 3%, and Broadcom also traded lower. The iShares Semiconductor ETF declined 1.2%, and the VanEck Semiconductor ETF lost 0.8%.

    Meta Rallies on AI Cost Report

    Meta Platforms stood out among large technology stocks, rising nearly 6% on Friday and extending its five-session advance to more than 14%. The move placed Meta on track for its strongest weekly gain since February 2024. Bank of America maintained its buy rating after reviewing details tied to Meta’s AI infrastructure plans. According to Reuters, Meta plans to begin production of an in-house AI chip in September as part of a broader plan to double computing capacity. “Meta may have engineered significant cost savings to get capacity cost per MW well below our and Street expectations,” Bank of America analyst Justin Post wrote.

    Circle Shares Jump After OCC Approval

    Circle Internet Group shares surged more than 12% after the stablecoin issuer received final approval from the U.S. Office of the Comptroller of the Currency to establish a national trust bank. The approval moves Circle further into the regulated U.S. banking system, supporting its USDC infrastructure and adding federally regulated custody. Reserve management is planned as a future capability.

    Economy and Stocks Move at Different Speeds

    The stock market’s performance continues to diverge from the broader economy. The S&P 500 rose nearly 10% in the first half of 2026, while the Dow gained almost 9% over the same period. Meanwhile, economists say U.S. growth has slowed. “I think there’s this widespread perception the two should be in sync,” said Joe Seydl, senior markets economist at J.P. Morgan Private Bank. “But, from a purely analytical perspective, they’re two very different phenomena.” Mark Zandi, chief economist at Moody’s, added, “We’re growing. We’re not in recession. But we’re not going anywhere quickly.”

    For Friday’s session, investors focused on weekly gains, chip stock weakness, Meta’s rally, Circle’s bank approval, and the first trading day for SK Hynix in the United States.

  • Meta to Begin Production of Custom Iris AI Chip in September, Targets 14 GW Computing Capacity

    Meta to Begin Production of Custom Iris AI Chip in September, Targets 14 GW Computing Capacity

    Meta Platforms has announced plans to begin production of its in-house artificial intelligence chip, code-named Iris, in September as part of a broader strategy to expand custom hardware development and reduce reliance on external suppliers. The move, outlined in an internal company memo reviewed by Reuters, aims to boost Meta’s computing capacity to 14 gigawatts next year while supporting the growing demands of AI workloads across Facebook, Instagram, and other services.

    Iris AI Chip: A Step Forward for Meta’s MTIA Program

    The Iris chip is part of Meta’s MTIA (Meta Training and Inference Accelerator) program. According to the memo, the chip completed bug testing in just six weeks with no major issues—a notable achievement for a program that has faced earlier delays since its launch over five years ago. Broadcom is assisting with design, while Taiwan Semiconductor Manufacturing Co (TSMC) is expected to handle production. The chip will not immediately replace GPUs from Nvidia and AMD but will complement them, supporting large-scale AI training and inference workloads.

    The memo notes that adopting the latest GPUs at Meta’s scale “has been a heavy lift, and it has cost us time,” underscoring the company’s push for greater control over its AI hardware stack. Meta declined to comment on the Reuters report.

    Computing Capacity Expansion: 14 GW by Next Year

    Meta plans to deploy 7 gigawatts of computing infrastructure this year, having added 1 gigawatt in the first half and expecting to add another 5.5 gigawatts by year-end. The company then aims to double total capacity to 14 gigawatts next year. To put that in perspective, 1 gigawatt can power approximately 800,000 homes, highlighting the immense energy requirements of large-scale AI data centers.

    This expansion comes as major tech firms ramp up AI infrastructure spending. Meta expects to spend up to $145 billion on AI infrastructure this year, part of a projected $700 billion in Big Tech spending on AI technology.

    Reducing Supplier Dependence with Custom Silicon

    Meta’s custom chip initiative is designed to reduce cost pressures tied to purchasing advanced AI chips from external suppliers like Nvidia and AMD. It also enables faster hardware iteration: Meta unveiled Iris under its technical name in March, along with three other AI processors, and plans to release a new chip roughly every six months through 2027—a much faster cycle than the industry standard of one year or more.

    Mike Gualtieri, a vice president and principal analyst at Forrester, commented, “You can’t become an AI titan if you are dependent on another company for chips.” He added that hyperscalers and SpaceX are also pursuing custom chip designs as model usage becomes more price-driven.

    Long-Term Supply Deals Support Data Center Growth

    To support its data center buildout, Meta has secured multi-year supply agreements with Samsung Electronics for memory chips, Sandisk for flash storage, and Sumitomo Electric for fiber-optic equipment. These deals are critical as AI demand strains the global chip supply chain, driving up prices for memory, storage, and AI processors. Sandisk declined to comment, while Samsung and Sumitomo Electric did not respond to Reuters requests.

    Morgan Stanley analysts have described rising chip and memory prices as “chipflation,” a broader concern for the technology sector. Meanwhile, companies like Apple have raised prices on some products amid memory supply constraints.

    Meta shares initially fell after the Reuters report but later recovered, trading up 4.6% in late afternoon trading after the company announced developer access to an AI coding model that competes with tools from OpenAI and Anthropic.