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).
🤖 Colleague Skill goes viral in China as workers clone colleagues into AI agents
An open source project called “Distill your colleague into an AI Skill” picked up ~9k GitHub stars in a few days.
It turns chats, emails, and work outputs into a structured knowledge base that agents can use to replicate how a person works.
The system captures how someone solves tasks, what tools they use, and how they communicate, then packages it into reusable “skills” for AI agents.
Chinese media report employees using it to replicate coworkers’ roles, aiming to stay replaceable-proof during layoffs.
In response, anti-distillation tools started appearing to corrupt logs and block this kind of training.
📊 @tech
📣 Milla Jovovich releases MemPalace, an AI memory system with 100% LongMemEval
An open source project from Milla Jovovich and Ben Sigman just dropped on GitHub. MemPalace claims 100% on LongMemEval, a level no model or agent has reached.
The system turns conversations into structured knowledge. It extracts facts, organizes them into a hierarchy, and uses semantic search to retrieve them. The approach is based on the “memory palace” method.
A key part is AAAK compression, which packs the knowledge base into about 120 tokens of context.
📊 @tech
⚡️ A good example of Tesla’s Automatic Emergency Braking reacting to a complex road situation
📊 @tech
🤖 OpenAI proposes rebuilding the economy for the AI era
OpenAI published a policy paper arguing current economic systems are not ready for AI. Sam Altman said the shift toward superintelligence is close and could reshape jobs, security, and social stability.
The plan treats AI as core infrastructure, with universal access similar to electricity. It suggests shifting taxes from labor to capital, since automation could shrink wage-based income.
It also proposes a public wealth fund to distribute AI-driven profits to citizens, and separating social benefits from employment as job stability declines.
Additional measures include audits of advanced models, global safety coordination, and testing a 4-day workweek.
Full paper: https://openai.com/index/industrial-policy-for-the-intelligence-age/
📊 @tech
⏰ Amida drops NASA Tribute Digitrend limited to 100 pieces
Amida released the Digitrend NASA Tribute to mark 50 years of the model. The run is limited to 100 watches, priced at 3,400 CHF.
Digitrend is built as a driver’s watch with a perpendicular display for reading time without moving your wrist. It uses a jumping hour and sliding minutes, powered by a 9-component mechanical system that mimics a digital display.
The design references the Space Shuttle era with a red NASA logo and a white ceramic top inspired by heat shield tiles.
📊 @tech
🤖 Anthropic cuts Claude subscriptions from third-party tools like OpenClaw
Since April 4, Anthropic disabled using Claude subscriptions inside external services. OpenClaw still works with Claude accounts, but usage is now billed separately via API token pricing.
The change targets heavy usage cases like large contexts and agent workflows, which exceeded the fixed $20 subscription model.
Workarounds rely on changing how requests are labeled so they look like first-party usage. Tool names get rewritten to pass filters.
📊 @tech
There are 10,136 active Starlink satellites at this moment
📊 @tech
📱 Apple bans vibe-coding apps as developers move them into iMessage
Apple removed apps like Replit, Vibecode, and Anything from the App Store, citing rules against executing code that changes app behavior after installation.
These tools allowed users to generate and run apps directly on iPhone without traditional development, which conflicts with App Store policies.
Developers behind Anything pushed an update that moves the functionality into iMessage, letting users build iOS apps without installing the main app.
Restrictions at the store level are pushing dev tools into alternative entry points.
📊 @tech
🤖 Anthropic targets 8,100 GitHub repos over Claude Code leaks
Anthropic sent DMCA requests against around 8,100 repositories hosting copies or forks of Claude Code. GitHub removed most of them within hours.
Some takedowns hit modified or unrelated projects. After уточнение требований, only 96 repositories remain under action, while the rest were restored. Projects like openclaude are still accessible.
Claude Code is Anthropic’s developer tooling tied to its models. The company is now actively enforcing control over its distribution.
📊 @tech
🚀 NASA launches Artemis II toward the Moon after 54-year gap
NASA’s Space Launch System has launched the Artemis II mission to the Moon.
It marks the first crewed lunar flight in 54 years.
More than 3 million people watched the launch live.
First images from the mission are expected in 6 days.
📊 @tech
📱 Telegram adds built-in AI text editor with Cocoon models
Telegram rolled out an AI editor that appears after typing a few words. The “Ai” button lets users translate text, fix grammar, or change style. It runs on open source models via Cocoon Network.
The feature sits directly in the input field and works as a quick rewrite tool inside chats. No separate app or workflow.
The update also brings media attachments to polls, Live Photos support on iOS, a bot builder, a document scanner on iOS, plus labels for unofficial clients.
📊 @tech
⛏️ Global mineral output surges as demand accelerates
A Reddit chart tracking global mining from 1960 to 2024 shows sharp growth across key resources. The data matches Our World in Data, based on USGS and BGS sources.
Lithium stands out at 240K tons in 2024, with near-vertical growth after 2015 driven by EVs and batteries. Cobalt reached 290K tons, and rare earths hit 390K tons, both rising fast on battery and electronics demand.
Copper and bauxite grew more steadily but at massive scale, with 23M tons of copper and 450M tons of bauxite in 2024.
Demand for minerals is rising at an exponential pace, especially for those tied to the energy transition.
The push to save the planet is driving deeper extraction of it.
📊 @tech
🤖 Claude Code gets computer use on macOS
Anthropic added Computer use to Claude Code. The agent can now write code, compile it, run the app, click through the UI, find a bug, fix it, and check the result from one prompt.
The feature is available as a research preview on Pro and Max plans, only on macOS for now. It can be enabled through /mcp in settings.
This pushes coding agents one step closer to handling the full loop, not just the code itself.
📊 @tech
🖥 Midjourney engineer open-sources Pretext for layout without CSS
A Midjourney engineer released Pretext, an open-source algorithm that handles text layout without CSS, DOM, or browser reflow. It computes layout directly with math instead of relying on traditional rendering pipelines.
This removes cascading styles and dependency chains, which often trigger heavy recalculations when content changes. The result is up to 500x faster layout in dynamic scenarios.
The approach fits emerging interfaces where content is generated on the fly, especially agent-driven UIs that rebuild structure step by step.
Layout is starting to move from browser logic into direct algorithms.
github.com/chenglou/pretext
📊 @tech
🌐 Samsung brings its browser to Windows with Perplexity inside
Samsung has renamed Samsung Internet to Samsung Browser and launched a full Windows version, for now only in South Korea and the US. The release adds deeper AI features on top of its existing mobile browser.
One feature is scroll sync. The browser opens pages on desktop at the exact paragraph where the user stopped on mobile.
Another is a built-in Perplexity layer that works with the current page, lets users search through history using natural language, and compare or search across open tabs.
Samsung is extending its browser across devices and embedding AI directly into browsing workflows.
📊 @tech
🔍 Google launches free AI voice recorder with offline mode
Google released AI Edge Eloquent, a free iOS app that turns raw speech into clean text. It works in real time, removes filler words, and copies the final transcript after recording. No subscription required.
The app runs fully offline, keeping data on-device. With Gemini enabled, it adds smarter processing through the cloud. It also supports summaries, tone editing, and tracks speaking stats.
Transcriptions are stored in history, and users can add custom words for better accuracy. It is part of Google’s AI Edge lineup focused on local AI tools.
📊 @tech
🤖 Anthropic launches Claude Mythos AI model but limits access
Anthropic introduced its new AI model, Claude Mythos, designed to detect vulnerabilities faster than humans. The model has already identified critical bugs, including a 27-year-old flaw in OpenBSD and issues in FFmpeg and Linux.
Access to Claude Mythos is restricted to major partners such as Amazon, Google, Microsoft, and Nvidia. Anthropic considers the technology too powerful for public release and is currently testing it within a closed security program.
📊 @tech
📺 Sber releases GigaChat-3.1 AI-models with open weights
Sber’s RnD team published GigaChat-3.1-Ultra and Lightning under MIT license on Hugging Face.
Both models are trained from scratch on Sber’s own compute and data, not fine-tuned from existing systems.
Ultra is a 702B MoE model that outperforms DeepSeek-V3-0324 and Qwen3-235B in math and reasoning benchmarks.
Lightning is a compact 10B MoE (1.8B active) that reaches high-level performance on arenas while staying fast and efficient.
The models support FP8 training, long context (up to 256k), and can run from large clusters to local environments.
It is clear that Sber is committed to making AI more accessible to developers by helping independent teams build assistants, copilots, and production systems on an open infrastructure.
📊 @tech
🤖 Users on X noticed Claude Code now throws an error when asked to analyze or work on Claude Code itself.
📊 @tech
📱 Netflix releases VOID, an open model for removing objects from video
Netflix published VOID, a model that removes objects from video while rebuilding scene physics. If a person holding a guitar is removed, the guitar falls naturally instead of staying frozen.
The system combines components from CogVideoX by Alibaba, Gemini 3 Pro by Google, SAM2 by Meta, plus data from Adobe. In blind tests, VOID was preferred in 64.8% of cases, while Runway scored 18.4%.
Available on Hugging Face: https://huggingface.co/netflix/void-model
📊 @tech
🌍 NASA releases Orion photos of Earth with auroras and zodiacal light
NASA published the first images from Orion during Artemis II. Commander Reid Wiseman captured Earth through the spacecraft window after the burn toward the Moon.
The photo shows two auroras and zodiacal light while Earth blocks the Sun.
It was shot on a Nikon D5 with f/4.0, 1/4s, 22 mm, ISO 51200. Resolution is 5568 × 3712, just over 20 MP.
📊 @tech
📱 Apple opens CarPlay to AI apps as ChatGPT launches on iOS 26.4
Apple added a new app category in CarPlay with iOS 26.4, allowing voice AI assistants. On March 31, OpenAI released ChatGPT for CarPlay as the first major app.
The interface is fully voice-based. No text or visuals on screen. Users open the app from the car display, speak, and hear responses. It mirrors the iPhone voice mode but adapted for driving. Activation is manual via the icon.
Support for other assistants like Claude and Gemini was reported earlier in February.
📊 @tech
🤖 Claude has announced that computer use in Claude CoWork and Claude Code Desktop is now available on Windows.
📊 @tech
🚕 100+ Baidu robotaxis stop mid-ride after system failure in China
More than 100 Baidu robotaxis suddenly halted during trips. Some stopped on busy roads, leaving passengers unable to exit safely.
The vehicles run as part of Baidu’s autonomous ride service. The incident shows how a single failure can affect many cars at once.
A similar case happened in December when Waymo vehicles stalled during a major power outage in San Francisco.
📊 @tech
Robert Greene will completely shift your perspective on how you use AI.
📊 @tech
🤖 Claude Code leak reveals roadmap for autonomous agents
Anthropic accidentally exposed Claude Code source via an npm package, where a source map file allowed full recovery of the codebase.
The leak spread quickly and was confirmed as a real internal CLI platform with ~500K lines of TypeScript.
Inside the code and related materials, future features and architecture details became visible:
🔸 Autonomous agents that run tasks in the background without user prompts
🔸 Proactive agents that can initiate actions on their own
🔸 Multi-agent coordination system for orchestrating complex workflows
🔸 “Auto” mode to execute actions without constant permission requests
🔸 Voice interface support, including speech recognition
🔸 Early work on memory, planning, and “sleep” cycles for agents
🔸 Internal system prompts and structure of agent reasoning
🔸 Dedicated coordinator module for managing multiple agents
https://github.com/instructkr/claude-code
📊 @tech
🤖 RAI shows Roadrunner robot that rolls and climbs stairs
RAI unveiled Roadrunner, a 15 kg robot that moves on wheels, climbs stairs, and balances on a single wheel. It switches between rolling and stepping depending on terrain.
The legs are fully symmetrical, with joints that bend in both directions. Wheels can be placed side by side or inline, and can lock for walking up steps.
RAI was founded in 2022 with $400M from Hyundai and is led by Boston Dynamics founder Marc Raibert.
The project is still a tech demo, focused on mobility and control systems.
📊 @tech
OpenClaw generates video on Seedance 2, import into premiere pro and edit.
All by itself.
📊 @tech
🧪GitLab founder uses AI to design his own cancer treatment
GitLab founder Sytse Sijbrandij built a personalized treatment approach after a rare spinal cancer returned and standard options were exhausted.
He gathered full datasets on himself, including genetics, scans, and medical records, and used AI to structure information, generate hypotheses, and find relevant research.
The process became an iterative testing loop, where multiple treatment ideas were explored in parallel and adjusted over time, similar to an R&D pipeline.
He is now in remission. AI acted as a tool to organize and accelerate decisions alongside resources and access.
Read more: https://sytse.com/cancer/
📊 @tech