3297625
A hub for startup news, trends, and insights, covering the global startup ecosystem for founders, investors, and innovators. Community: @startupdis Buy Ads: @strategy (this is our only account).
🤖 Claude Code creates live pages from sessions
Anthropic added artifacts to Claude Code, letting users turn any work session into a live web page with code, chat history, and tool data all in one place. The page updates automatically at the same URL and keeps version history.
This feature supports scenarios like PR reviews, incident timelines, license audits, and architecture overviews. Pages are private by default, visible only to authorized organization members. It is currently in beta for Team and Enterprise plans.
📊@tech
A scam operation targeting men online was set up in China using AI.
One phone types messages with a robotic hand while another guides where to tap via camera.
The AI pretends to be a woman, flirts with men, and extracts money.
The scheme mimics real human chat.
📊@tech
😺 Codex adds Record & Replay feature
Codex introduced Record & Replay, a feature that lets users teach an agent to perform tasks on a computer by showing the process once.
The agent learns sequences like uploading videos to YouTube, editing photos, or signing documents, then repeats them automatically.
The feature is rolling out gradually to Codex app users.
📊@tech
🤖 Anthropic updates Claude Design with brand style sync
Anthropic's Claude Design now integrates your design system directly from repositories, design files, or codebases to maintain your brand style across projects. It builds interfaces using your real components and checks compliance before you see the result.
The editor is more stable for daily use and adds new layout controls. You can drag, resize, and align elements on the canvas without extra steps. Claude Design and Claude Code sync bidirectionally, letting you start in code or design and keep projects aligned.
Finished work exports to PDF, PowerPoint, or other tools you already use. This update tightens the workflow between design and development with fewer style mismatches.
📊@tech
🛰 Midjourney launches full-body ultrasound scanner
The Midjourney Scanner uses a ring of thousands of sensors and 40 ultrasound chips from Butterfly Network to capture 3D images of muscles, fat, bones, and organs in about a minute without radiation or magnets.
Users stand on a platform that lowers them into a shallow pool surrounded by the sensor ring. It creates detailed body composition maps comparable to MRI quality. About a dozen people have been scanned so far.
Midjourney plans to install ten scanners in a San Francisco spa opening by the end of next year. The spa will offer body scans alongside saunas and pools. FDA approval is pending, so use is limited to body composition imaging.
The connection between Midjourney’s AI image generation.
📊@tech
🤖 Boston Dynamics Atlas masters new tasks
Boston Dynamics' humanoid Atlas adapts to new environments and tasks beyond pre-programmed actions. Simulations equal to millions of training hours run daily, with skill transfer to the robot in about an hour, per KB Securities.
Atlas now moves a refrigerator over 45 kg, despite training on 23-32 kg loads, using full-body coordination and balance to adjust force dynamically.
Engineering changes simplify mechanics: two actuator types, symmetrical limbs, and continuous-rotation joints without cables. This reduces the simulation-to-reality gap and improves durability.
An analyst predicts Boston Dynamics could capture up to 60% of the premium industrial humanoid market by 2035.
📊@tech
🔗 MiniMax launches AI-powered creative Hub
MiniMax introduced Hub, a local workstation where AI agents handle creative tasks like research, scripting, image creation, music, and final editing. It supports multiple projects and an infinite canvas for seamless workflow.
Hub runs directly on your computer, integrating local files and apps for instant import and export. Early users before July 1 get 3000 credits, and points from Hailuo AI transfer fully to MiniMax Hub.
📊@tech
🤖 Anthropic sued over subscription limits
A US user named Carl Khan filed a lawsuit accusing Anthropic of misleading customers by lowering subscription limits. He claims the advertised limits for premium plans Max 5x and Max 20x are far from reality.
For example, the Max 20x plan reportedly offers only about 6× the limits of the Pro subscription, not the promised 20×. Khan wants to make the lawsuit class-action, covering all users who bought these plans since April 2025.
📊@tech
🍏 Apple Holds Back 3 iOS 27 Features
Apple plans to release three unannounced features with the iPhone 18 Pro and new Apple Watch this fall. These include a new watch face, a redesigned Camera app, and expanded Siri support for third-party chatbots.
The new watch face is likely a simplified Modular Ultra design, bringing Ultra aesthetics to standard Apple Watch models. The Camera app update will let users customize which functions appear and in what order during shooting.
Siri will soon support multiple AI assistants beyond Apple’s own, including ChatGPT, Gemini, and Claude. This effectively turns Siri into a platform where users choose the AI handling their voice requests.
📊@tech
🤖 OpenRouter launches Model Fusion system
OpenRouter introduced Model Fusion, a system that merges multiple neural networks into one advisory unit. It supports combinations like GPT-5.5 with Opus 4.8 tailored for specific tasks.
These model blends can outperform Claude Fable 5 in efficiency while running faster and costing less.
📊@tech
The first commercial computer was unveiled on this day in 1951.
It weighed 16,000 pounds.
Smartphones today weighs less than half a pound.
And are millions of times more powerful.
📊@tech
🔍 Google Gemini 3.5 translates live speech in 70 languages
Google launched Gemini 3.5 Live Translate, which translates live speech into over 70 languages without pausing. The model processes speech in real time, preserving tone and pace, and marks each synthesized segment with a SynthID watermark.
The feature expanded Google Meet’s language support from 5 to more than 70 languages and over 2000 language pairs. Developers can access it via Gemini Live API, while users get it in the Google Translate app on Android and iOS.
📊@tech
🤖 US Orders Anthropic to Block Mythos and Fable
The US government has ordered Anthropic to cut off access to Mythos and Fable for all foreign nationals, including those inside the US and foreign employees of Anthropic.
Anthropic apologized but complied immediately, shutting down access to these models. They called it a misunderstanding and are working to restore access as soon as possible.
📊@tech
✈️ Telegram launches native app on smartwatches
Telegram is now on all major smartwatches, including Apple Watch, Samsung, Google, and Xiaomi.
Users can read messages, send stickers, record voice notes, and share location from their wrist.
📊@tech
🤖 Anthropic's Claude Fable 5 Outperforms Peers
Real tests show Claude Fable 5 beating Claude Opus 4.8, Gemini 3.1 Pro, and GPT 5.5 on tough visual tasks like 3D hydrodynamics and complex physics. The gap is significant.
Fable 5 is Anthropic's most powerful model open to all users. It is available now and free on Pro, Max, Team, and Enterprise plans until June 22.
📊@tech
🤖 ChatGPT beats doctors in medical accuracy
OpenAI reports that ChatGPT now answers medical questions with 71% fewer errors over two months. This improvement comes from manual review of 700,000 responses by 260 doctors across 60 countries.
GPT-5.5 Instant powers ChatGPT, which is free for users. It handles 230 million health questions weekly, but all accuracy tests were done internally without independent clinical validation.
📊@tech
📺 Amazon Drops Nearly Finished Sam Altman Film
Amazon abruptly refused to release the almost completed film Artificial about Sam Altman's 2023 firing. The movie stars Andrew Garfield as Altman and Yura Borisov as Ilya Sutskever.
The decision came right after Amazon signed a $50 billion deal with OpenAI. They gave no clear reason, only saying the film would suit another distributor better.
📊@tech
🤖 ChatGPT adds centralized task scheduling
OpenAI introduced a new Scheduled page in ChatGPT that shows all active tasks with start times and control buttons. Users can set tasks for specific times or general periods like "morning" or "evening."
ChatGPT now monitors selected websites and connected apps for changes and notifies users. This replaces the Pulse daily digest feature. The update is available for Plus, Pro, Business, and Enterprise plans.
📊@tech
💰 Still paying up to 3% per transaction?
Traditional payment processors cut into your margins with high fees, rolling reserves, chargebacks, and slow settlements.
🔥 B2BINPAY helps businesses accept crypto while settling directly in fiat, securely and efficiently.
What you get:
🖱 Fees from 0.25%
🖱 0% rolling reserve
🖱 No chargebacks
🖱 Fast settlement
🖱 350+ digital assets accepted (BTC, ETH, USDT & more)
🌟 Built for business:
Regulated VASP | Audited security | $5.1B+ processed | 24/7 support
❕ Protect your margins & get started with B2BINPAY - Register today
👉 Join the B2BINPAY Telegram channel
Stop letting payment providers control your cash flow. Upgrade to payment infrastructure built for growth.
Disclaimer: Service has legal and jurisdictional limitations. Check T&Cs here
🔍 Google launches smart speaker with Gemini
On June 25, Google will start selling the Google Home Speaker for $100, its first smart speaker since 2020. It features Gemini, a new assistant replacing the old one.
Gemini enables full conversations and integrates with third-party services, moving beyond simple commands like playing music.
📊@tech
🚀 SpaceX Unveils AI1 Satellite Data Center
SpaceX introduced AI1, the first satellite data center for AI tasks, per Bloomberg. Elon Musk says AI1 will join a future network of about 1 million satellites performing AI computing in low Earth orbit.
The satellites will have solar panels up to 70 meters long and support 150 kW computing loads with Nvidia chips. Musk says building them is simpler than Starlink satellites, needing mainly solar panels, cooling, and laser links.
SpaceX plans a major production expansion in Texas. The new Gigasat facility will span over 11 million sq ft (~1 sq km) to make large solar panels for these space data centers.
📊@tech
📣 SpaceX to Merge with Cursor for $60B
SpaceX and Cursor signed a final merger agreement valuing Cursor at $60 billion. The deal will be done through a stock swap and aims to close in Q3.
Cursor is a coding tool with a loyal user base, unlike SpaceX's Grok, which failed to gain traction in vibe coding and had an unnoticed Grok Build release.
📊@tech
🔍 DeepMind Explores Life Beyond AGI
DeepMind published a detailed study titled "From AGI to ASI" arguing that AGI is not the ultimate goal but just the start. They stress that human-level AI is limited by biology and marketing has oversold it as a finish line.
AGI means Artificial General Intelligence, AI at human level. But human intelligence is specialized and bounded by evolution. Machines can scale memory and speed far beyond human limits.
The real target is ASI, Artificial Superintelligence, which surpasses entire organizations. This could come from scaling many AGI instances or new breakthroughs. Even millions of AGI copies running simultaneously could count as ASI.
📊@tech
👟 Adidas makes World Cup ball a smart device
The official World Cup ball Trionda must be wirelessly charged before each match. It contains a 14-gram motion sensor operating at 500 Hz, tracking position, speed, spin, and direction 500 times per second. A full charge lasts about 6 hours, enough for the entire game including extra time.
Sensor data combines with stadium cameras to create a real-time 3D model of every play. This feeds the VAR room to speed up and improve decisions on offsides, handballs, and goal-line calls. At around $170, Trionda may be the most expensive standard match ball in World Cup history.
📊@tech
🌱 Tric deploys UV robots that cut pesticide use by 70%
Tric is deploying autonomous robots that move through fields at night, using UV-C light to kill fungal diseases and spider mites without chemicals. Pilots on California's Central Coast have shown up to 70% reductions in pesticide use.
The technology is based on a simple biological advantage. Many fungi can repair UV damage when exposed to blue light during the day. At night, that repair mechanism does not activate, making the damage permanent. University trials found nighttime UV treatments could eliminate up to 95% of infections on crops including strawberries, grapes, cucumbers, basil, and roses.
⠀
Tric's Eden robot spans more than 40 feet, while its Luna platform covers 6 rows at a time across 50 to 100 acres. The machines can carry UV systems, bug vacuums, and precision cameras in a single pass.
⠀
Farmers do not buy the robots. They subscribe to the service, while Tric operates the fleet.
📊@tech
🍏 Pixar Teases New Film Gatto
Pixar teased Gatto, about a cat named Nero who owes the Venetian mafia. Directed by Enrico Casarosa of Luca.
The animation mimics Puss in Boots 2 with a 2D drawing style.
It will release in March next year.
📊@tech
🤖 OpenAI and Anthropic Subscriptions Unprofitable
SemiAnalysis tested OpenAI and Anthropic subscriptions by running long tasks to use weekly limits. The $20 OpenAI plan delivers about $700 in token value, the $200 plan about $14,000. Anthropic’s $100 plan shows $2,000 in token costs, revealing heavy losses across tiers.
Both companies offer fixed-price subscriptions with variable token usage depending on tasks.
OpenAI’s plans lose money below 15% usage, Anthropic below 20%. The priciest plans show the worst margins: ChatGPT-pro-20x hits -775% at 50% load, Claude-max-20x -400%. This may limit startups to API-only access instead of adding new subscription features.
📊@tech
🫡 Xiaomi launches free Claude Code alternative
Xiaomi released MiMo Code, a free terminal AI agent for programming built on OpenCode with an MIT license. It handles long tasks of hundreds of steps by saving state to files and restoring context seamlessly, avoiding memory loss common in other agents.
MiMo Code divides memory into four levels: session notes, project knowledge, user settings, and full action history. It cleans duplicates weekly and converts repeated actions into skills monthly. A Max Mode runs five solutions in parallel, boosting benchmark scores by 10-20% but increasing token use 4-5 times.
📊@tech
🤖 Anthropic reverses hidden AI limits after backlash
Anthropic changed its policy less than 48 hours after releasing its AI model Fable. The startup faced criticism for secretly degrading responses to queries related to AI development without notifying users.
Originally, Anthropic openly redirected chemistry, biology, and cybersecurity questions to a weaker model, Opus 4.8. However, a hidden rule quietly reduced answer quality for any requests resembling AI distillation or development attempts.
Engineers received no alerts when these limits triggered, leading to confusion and accusations of sabotage. After public outcry, Anthropic apologized and now clearly informs users when requests are refused or downgraded to reduce misuse risks.
📊@tech
🔍 Google Gemini 3.5 Live Translate launched
Google introduced Gemini 3.5 Live Translate, an audio model that delivers instant voice translation in over 70 languages. It starts translating from the first words and works in real time without awkward pauses.
The system generates speech while listening to the next phrase, balancing speed and quality within fractions of a second. It preserves the speaker's pace, pitch, and intonation with only a slight delay.
Gemini 3.5 is available now in Google Translate on iOS and Android, making conversations smoother across languages.
📊@tech