Search the archive

ROBOSTRAL NAVIGATE: SINGLE-CAMERA AI NAVIGATION

Mistral's Robostral Navigate is an 8B AI model that enables robots to autonomously navigate environments using a single RGB camera.

META'S NEW MUSE IMAGE MODEL CAN PULL OTHER INSTAGRAM USERS INTO AI PHOTOS

Meta's Superintelligence Labs launched Muse Image, its first AI image model, now powering image tools in the Meta AI app, Instagram, and WhatsApp, with Facebook and Messenger coming soon. The model is "agentic," reasoning through prompts and searching the web before generating, and lets users @ mention Instagram accounts to incorporate their public photos into AI-generated images. Muse Image also enables room redesigns from Marketplace photos, direct drawing edits on images, and powers 30 new AI effects rolling out to Instagram Stories in the US.

TRIPO AI RAISES $150M FOR GENAI TOOLS FOR GAMING — A MONTH AFTER ITS PREVIOUS $200M RAISE

Tripo AI raised $150 million in a Series A3 round, just a month after closing $200 million across previous rounds. Investors spanned automotive, gaming, and tech sectors, including Geely Capital, 4399 Network, Tanwan, Giant Network, Fosun Capital, and others. The funding will support further development of its 3D foundation models and Project Eden world model, plus global commercialization efforts.

APPLE OVERHAULS RAW PHOTO PROCESSING WITH IOS 27, SHOWCASES IMPRESSIVE RESULTS

Apple is introducing RAW 9 in iOS 27, iPadOS 27, and macOS 27, its biggest upgrade yet to the system-level RAW image processing engine. Powered by on-device machine learning and the Apple Neural Engine, RAW 9 combines demosaicing and noise reduction to deliver sharper details, more accurate colors, and significantly better low-light performance—even when reprocessing older RAW photos from nearly 800 supported camera models.

WHEN BRAND GUIDELINES RUIN YOUR WEBSITE (AND HOW TO FIX IT)

Rigid brand guidelines often prioritize visual polish over web usability, quietly hurting conversion through poor contrast, unreadable type, and missing hierarchy. Testing a brand-compliant homepage against a tweaked version through an AI attention tool showed clarity and focus up ~20% and predicted engagement with the main call to action up over 80%. Rather than treating brands as sacred, businesses should test whether their branding actually communicates intended values and works on screen, fixing readability conflicts before considering a full digital-first rebrand.

BUILDING ACCESSIBILITY INTO A CANVAS-BASED PRODUCT

Figma built a synthetic "Mirror DOM" to bring screen reader and keyboard accessibility to its canvas-based renderer, which otherwise bypasses the browser's native accessibility tree entirely. The system relies on an internal accessibility tree that generates per-layer summaries, a recursive React component that mirrors those layers into hidden but spatially-positioned DOM elements, and bidirectional syncing between canvas selection and system focus. This work expanded from Prototype viewer support in 2022 through FigJam and Design/Dev Mode/Slides improvements in 2024-2025, developed with accessibility feedback from Fable's assistive-technology users.

FIGMA'S LOREDANA CRISAN TALKS TASTE, LAST-MINUTE DECISIONS AND THE ART OF PRIORITISATION

At Config 2026, Figma unveiled new AI-powered features while reinforcing its belief that AI should enhance—not replace—human creativity. Chief Design Officer Loredana Crisan argued that great design comes from making, experimentation, and human judgment, with AI serving as a tool that supports creative work rather than automating it. She also highlighted the importance of fostering a culture where ideas can gain momentum, designers embrace AI-driven experiences, and empathy remains the most valuable skill as the industry evolves.

LEVEL UP YOUR ACCESSIBILITY GAME

128 questions to level up your web accessibility game. Practice WCAG 2.2, ARIA, semantic HTML, keyboard, and contrast, with plain-English answers.

SHADER MAGIC FOR FRONTENDS

The component library for creative WebGPU effects in the browser.

EDIT ENTIRE VIDEOS WITH A SINGLE PROMPT

Drop in your clips and say what you want. The cuts, captions, and music land in minutes, ready to post.

10 AI SKILLS TO GIVE YOUR CODING AGENT REAL DESIGN TASTE

Design taste isn't something coding agents have by default — they need it packaged as installable rules, much like the ten markdown-based skills covered here, spanning color science, Swiss grid systems, and brand-specific style generation. Each skill teaches a distinct design language that agents apply automatically once installed in a project. Design knowledge is moving from static references humans read to executable rules agents run directly.

TRUMP'S PLAN TO REDESIGN EVERY .GOV WEBSITE LEADS TO AI-DESIGNED HORRORS

The National Design Studio, created by Trump's 2025 executive order to redesign 27,000 government websites in three years, has produced mostly single-page sites and faces mounting criticism a year in. Investigations found NDS sites used undisclosed commercial trackers, built unauthorized versions of vote.gov and passport.gov requiring sensitive identity verification, and shipped AI-generated errors like a six-toed child on TrumpRX.gov. Agencies are now resisting NDS outreach, and a July 4 deadline for new web design standards has quietly been dropped from the executive order.

WHY MAKING A BRAND FEEL LIKE ITSELF IS THE REAL CHALLENGE

Strong brand systems make visual consistency easier than ever, but many brands are losing their distinctiveness because their behaviour—personality, tone, pacing, and confidence—is harder to define and maintain than logos, colors, or typography. As content production scales across more channels and teams, consistent brand behaviour becomes a key differentiator, with motion design playing a particularly important role in expressing a brand's character and making it instantly recognizable.

WHERE DO ALL THE WOMEN GO? THE CREATIVE INDUSTRY'S QUIET TALENT DRAIN

Women with two decades of high-level creative experience are quietly leaving the industry once they reach the top, as rigid, presence-based work structures fail to accommodate their lives outside work.

11 AI PROMPTS EVERY UX DESIGNER SHOULD SAVE

AI can do more than speed up UX work.

REWRITING BUN IN RUST

Bun was initially developed as a port of esbuild's transpiler from Go to Zig, with a broad scope including transpilation, package management, and HTTP client capabilities. Following its acquisition by Anthropic and to address stability issues linked to Zig's manual memory management, Bun has been rewritten in Rust, using Rust's safety features like automatic memory cleanup to improve stability and reduce bugs.

SEPARATING SIGNAL FROM NOISE IN CODING EVALUATIONS

A recent audit by Anthropic showed that approximately 30% of tasks in SWE-Bench Pro, a widely used coding evaluation benchmark, are broken due to issues such as overly strict tests, underspecified prompts, and misleading instructions. The analysis involved a combination of automated screening and human reviews by experienced software engineers. Given the flaws discovered, a recommendation to transition to SWE-Bench Pro has been retracted.

A BUG WHICH ONLY AFFECTED LEFT-HANDED USERS

A bug in WordPress impacted only left-handed users, causing comment boxes to appear unexpectedly while scrolling, which the author found did not affect right-handed users like himself. The issue came from outdated code that had not been addressed in years, which was finally fixed by removing a few lines of code.

I THINK I HAVE LLM BURNOUT

Frequent use of LLMs for coding tasks has led to feelings of burnout for devs due to the repetitive and often flawed nature of their outputs, which include false assumptions and excessive stylistic quirks. Despite increased productivity and a learning curve, the monotony of LLM-generated content is becoming increasingly tiresome and frustrating.

EXPERIENCES WITH LOCAL MODELS FOR CODING

Recent experiences with locally run models for agentic coding reveal both challenges and successes, particularly influenced by task complexity and the machines used. The evaluations show that while these models can handle straightforward coding tasks effectively, their performance varies a lot.

ANNOUNCING TYPESCRIPT 7.0

TypeScript 7.0 introduces a large performance boost, achieving up to 10 times faster compilation through a native port built in Go. This release has been rigorously tested in real-world environments, leading to improvements in development efficiency and user experience across various large-scale projects and organizations.

INTRODUCING GROK 4.5

Grok 4.5 is the latest and most advanced model by SpaceXAI, designed to excel at coding, agentic tasks, and knowledge work.

INTRODUCING GPT-LIVE

GPT-Live introduces a new generation of voice models designed for more natural and interactive human-AI conversations, using a full-duplex architecture that allows for simultaneous listening and speaking. This advancement allows for smoother dialogue, better responsiveness, and the capability of delegating complex tasks to specialized models.

AN OFF SWITCH FOR DUAL USE KNOWLEDGE IN AI MODELS

AI models contain dual-use knowledge that can be used positively or negatively, such as cyber defense or exploitation. GRAM (Gradient-Routed Auxiliary Modules) allows a single model to have dedicated compartments for different categories of dual-use knowledge. Preliminary results show GRAM can effectively isolate and control dual-use capabilities while resisting attempts to recover removed knowledge.

INTRODUCING MEERKAT: AN EXPERIMENT IN GLOBAL CONSENSUS

Cloudflare Meerkat is a distributed consensus service designed to manage global control-plane state across its numerous data centers while ensuring strong consistency and fault tolerance. Unlike traditional consensus algorithms like Raft, which rely on a single leader, Meerkat uses the QuePaxa algorithm, allowing all replicas to write simultaneously and maintaining system availability even during failures.

GEOSQL

GeoSQL is a geospatial analytics tool that helps with data science and analysis workflows on platforms like PostGIS, BigQuery, and Snowflake by integrating a map-based validation loop to improve accuracy and performance of spatial queries.

CLOUDFLARE DROP

Cloudflare allows users to quickly upload and deploy HTML, CSS, and JS files to see their site live instantly.

OWNERSHIP

The concept of ownership in engineering means taking full responsibility for problem-solving from identification to implementation, which means detailed consideration of the problem, potential solutions, edge cases, testing, and communication.

GITLOST: HOW WE TRICKED GITHUB'S AI AGENT INTO LEAKING PRIVATE REPOS

Noma Labs uncovered a critical vulnerability named GitLost in GitHub's newly launched Agentic Workflows, which allowed unauthorized users to extract data from private repositories by exploiting an indirect prompt injection attack through public GitHub issues.

Apply now and skip the waitlist →

USER-CONTROLLED ALGORITHMS: SHOULD YOU CARE?

User-controlled feed settings should not change your core ad strategy because platforms still rely more on user behavior than declared interests to decide what people see. Create ads that hold attention and earn meaningful engagement because watch time, rewatches, and skips reveal more than interest categories. Test multiple creative formats and messages so the platform can identify which version resonates with each audience based on real viewing behavior. Use performance data to guide creative decisions instead of relying on the preferences users select in platform settings.

THE SAME ARTICLE IS RUNNING ON A HUNDRED LOCAL NEWS SITES. GOOGLE HAS NOTICED

Local news outlets are increasingly publishing syndicated articles created from public datasets, with the same stories appearing across hundreds of sites. While publishers get free content, syndicators gain backlinks, distribution, and brand partnerships by leveraging trusted mastheads. The model weakens local journalism by replacing original reporting with duplicate content that competes in search and erodes reader trust. As Google cracks down on duplicate and third-party content, outlets may need to prioritize original reporting and owned audiences over rented reach.

HUBSPOT REVERSES COURSE AFTER CUSTOMERS REJECT DATA SHARING

HubSpot's attempt to build a shared enrichment pool across customer accounts exposed a broader tension in how vendors profit from data flowing through their own platforms. Customers contribute business contact, engagement, and tracking data simply by using the product, so once that data becomes an input for a commercial dataset sold back to every other account, including competitors, the value exchange shifts without customers getting anything new in return. HubSpot's five enrichment controls only governed whether an account's own records got refreshed, not whether its data flowed into the shared pool, making participation effectively mandatory.

EMAIL COURSE > PDF LEAD MAGNET

Most lead magnets fail because PDFs get downloaded once and forgotten, missing the opportunity to build an email habit. Instead, turn the content into a 5-day email course that delivers each lesson, framework, and takeaway through daily emails. This keeps subscribers engaged and creates a natural transition into a 3-5 email pitch sequence that drives the next step.

WIN REPORT: HOW REVERSING THE FUNNEL INCREASED ORDERS BY 22%

Moving a customer's first moment of personal investment earlier in the purchase journey can increase conversions. A flower and gift retailer moved the card message step to immediately after product selection and orders increased by 22%. Writing a personal message creates commitment and helps customers picture the outcome before payment. Look for opportunities to let customers personalize or create something early in the funnel and test whether it increases purchase completion.

IS RUNNING THE SAME AI PROMPT ONCE A DAY ACTUALLY ENOUGH?

Analysis of one million AI responses found running prompts 10X daily offers minimal gains over once per day, with only a 0.25 point difference in citation share. While higher frequency slightly reduces variability, the impact is small. The bigger driver of accuracy is prompt selection, not repetition. A well-constructed, representative prompt portfolio matters more than run volume, since poor prompt choices can skew results far more than day-to-day AI response variation.

HOW AI-NATIVE ARE YOU AS A MARKETER? (QUIZ)

Built by CXL, this quick assessment gauges how AI-native you are as a marketer. It tracks how consistently you use AI across workflows, research, content creation, experimentation, and team operations. Your level reflects how broadly and often you apply these behaviors in your day-to-day work.

SEE HOW CONTENT FROM SOCIAL AND VIDEO PLATFORMS PERFORMS ON GOOGLE SEARCH

Google launched platform properties in Search Console, enabling creators to track how Instagram, TikTok, X, and YouTube content performs on Search and Discover. It surfaces queries, clicks, impressions, and engagement through performance and insights reports, with options to filter or export data. Users can connect accounts via verification, with a gradual rollout designed to give a unified view of content discovery across platforms.

WHY DOES THE MIDDLE OF THE FUNNEL ALWAYS GET THE MIDDLE FINGER?

CMOs now spend 71% of budget on performance media and just 29% on brand-building, with performance gaining 11 percentage points since 2023, moving further from the 60/40 optimal split. A comparison of premium and non-premium publisher placements found the gap was modest at the top and bottom of the funnel but reached +267% in the middle, the exact zone budgets are abandoning. Mid-funnel performance depends on media quality, specifically placement prominence and contextual receptiveness, neither of which shows up in reach and frequency or click and ROAS metrics.

EVERYONE'S A MARKETING EXPERT ... UNTIL THEY'RE NOT

Marketing's biggest blind spot is not inexperience but certainty, since years of success can calcify a personal lens that has quietly drifted out of alignment with the audience. Performance metrics like conversions and cost per acquisition measure whether a campaign worked, but they cannot explain why an audience felt connected or alienated in the first place. The marketers who avoid that trap treat their own reactions as a starting hypothesis rather than a final verdict, testing and adjusting based on what the data shows rather than what they personally like.

META'S NEW AI CAN GENERATE IMAGES OF YOU FROM YOUR INSTAGRAM, AND YOU'RE OPTED IN

Muse Image lets anyone @-mention a public Instagram account to generate AI images of that person, switched on by default with no consent request or notification.

WHOOP HIRES NIKE MARKETING VETERAN AS WEARABLES WARS HEAT UP

Former Nike CMO Dirk-Jan van Hameren joins Whoop as CMO.

EMAIL OPEN TRACKING NOW NEEDS CONSENT IN FRANCE AND ITALY

France's CNIL and Italy's Garante now require consent for email tracking pixels; deadlines July 14 and October 28, respectively.

Attend live on Tuesday 28 July

META ENTERS AI IMAGE MODEL RACE IN BID TO COURT ADVERTISERS AND SUBSCRIBERS

Meta has released a new AI model for creating images, Muse Image. The model will be available for consumers to access for free via the Meta AI app and site, WhatsApp direct messages, and Instagram Stories. Users who hit their free limit can purchase a subscription or wait until the limit resets. Muse Image will be used to power advertiser-specific image-generation tools as part of Meta's AI-powered Advantage Plus service.

BEHIND XBOX'S BIG LAYOFFS, A STREAMING STRATEGY THAT FAILED

Xbox announced on Monday that it would lay off 3,200 employees and let go of five game studios. The company is aiming to reset itself after overspending on studios, staff, and games that people didn't want to play. Xbox plans to take a more streamlined approach to games and studios going forward. It will focus more on franchises like Minecraft, which is considered one of the most successful video games in the world.

SPACEX JUST LAUNCHED THE 1ST-EVER NUCLEAR-POWERED COMMERCIAL SATELLITE

The BOHR (Betavoltaic Orbital High-Reliability) satellite is a commercially built nuclear-powered cubesat created by Florida-based company City Labs. It was designed to test the company's proprietary NanoTritium betavoltaic micropower source in space for the first time. The NanoTritium device harnesses beta particles emitted from the radioactive decay of tritium and converts them directly into electricity using a semiconductor. It is meant to provide continuous power to spacecraft without a reliance on solar energy.

TESLA CYBERCAB INCLUDES MORE POWERFUL FSD HARDWARE

Production units of the upcoming Cybercab are running a more powerful version of Tesla's Full Self-Driving computer compared to the Model 3 or Model Y. The biggest change is the increased onboard memory, which will allow the computer to run much larger self-driving models. The extra computing muscle enables the Cybercab to be equipped with an SAE Level 4 Autonomous Mode - the Cybercab is designed to operate entirely without human intervention. Tesla started testing Cybercabs without steering wheels or pedals on public roads in Austin last week.