OpenAI Launches Standalone ChatGPT Translate Tool
OpenAI has launched ChatGPT Translate, a dedicated web-based translation service that works separately from the main ChatGPT interface. It supports text translation in 50+ languages and is currently available via its own website.
The interface looks similar to other online translators, with fields to enter source and translated text, and dropdown menus to choose languages. On mobile browsers it also supports voice input, though some features like image translation aren’t live yet.
The tool isolates translation capabilities that were previously part of the ChatGPT chatbot into a standalone web experience, making it easier for users who just need translation without the full chat interface.
OpenAI has launched ChatGPT Translate, a dedicated web-based translation service that works separately from the main ChatGPT interface. It supports text translation in 50+ languages and is currently available via its own website.
The interface looks similar to other online translators, with fields to enter source and translated text, and dropdown menus to choose languages. On mobile browsers it also supports voice input, though some features like image translation aren’t live yet.
The tool isolates translation capabilities that were previously part of the ChatGPT chatbot into a standalone web experience, making it easier for users who just need translation without the full chat interface.
❤1
This media is not supported in your browser
VIEW IN TELEGRAM
“If you’re going to do anything new or innovative, you have to be willing to be misunderstood.”
~ Jeff Bezos
✔️@technco
~ Jeff Bezos
✔️@technco
This media is not supported in your browser
VIEW IN TELEGRAM
“If you’re going to do anything new or innovative, you have to be willing to be misunderstood.”
~ Jeff Bezos
~ Jeff Bezos
🚀 SpaceX files confidential IPO aiming for up to $75B raise
SpaceX has confidentially filed for an IPO, with plans to raise up to $75B. The company was recently valued at $1.25T after merging with xAI, with reports pointing to a potential $1.75T target at listing.
The company runs reusable rocket launches and operates Starlink with around 10,000 satellites. It has also secured over $24.4B in U.S. government contracts since 2008 and completed 165 orbital flights in 2025.
If completed, the IPO would be the largest in U.S. history, far exceeding Alibaba’s $22B record. Timing remains tied to market conditions, with volatility still elevated.
SpaceX has confidentially filed for an IPO, with plans to raise up to $75B. The company was recently valued at $1.25T after merging with xAI, with reports pointing to a potential $1.75T target at listing.
The company runs reusable rocket launches and operates Starlink with around 10,000 satellites. It has also secured over $24.4B in U.S. government contracts since 2008 and completed 165 orbital flights in 2025.
If completed, the IPO would be the largest in U.S. history, far exceeding Alibaba’s $22B record. Timing remains tied to market conditions, with volatility still elevated.
⚠️ Oracle cuts up to 30,000 jobs in one day to fund AI push
Oracle laid off around 10,000 to 30,000 employees in a single day, about 18% of staff. Emails went out at 6am with no prior notice, marking the same day as the last working day. Some divisions saw cuts above 30%.
The company is shifting spending toward AI and cloud infrastructure. In 2025, rising costs created a major cash gap, pushing management to reallocate capital.
Analysts at TD Cowen estimate the layoffs could save $8–10B in free cash flow. Even if fully reinvested into AI infrastructure, payback is not expected before 2030.
Oracle laid off around 10,000 to 30,000 employees in a single day, about 18% of staff. Emails went out at 6am with no prior notice, marking the same day as the last working day. Some divisions saw cuts above 30%.
The company is shifting spending toward AI and cloud infrastructure. In 2025, rising costs created a major cash gap, pushing management to reallocate capital.
Analysts at TD Cowen estimate the layoffs could save $8–10B in free cash flow. Even if fully reinvested into AI infrastructure, payback is not expected before 2030.
📉 OpenAI shares stall as $600M hits market with no buyers
Investors are struggling to sell around $600M worth of OpenAI shares on secondary markets. Deals that used to clear in days are now seeing no demand, according to Next Round Capital.
At the same time, marketplaces like Augment and Hiive report record demand for Anthropic shares. The gap between OpenAI’s $852B valuation and Anthropic’s $380B is pushing buyers toward Anthropic equity.
Banks including Morgan Stanley and Goldman Sachs have started offering OpenAI shares to clients without carry fees. Goldman still charges 15–20% carry on Anthropic deals.
Investors are struggling to sell around $600M worth of OpenAI shares on secondary markets. Deals that used to clear in days are now seeing no demand, according to Next Round Capital.
At the same time, marketplaces like Augment and Hiive report record demand for Anthropic shares. The gap between OpenAI’s $852B valuation and Anthropic’s $380B is pushing buyers toward Anthropic equity.
Banks including Morgan Stanley and Goldman Sachs have started offering OpenAI shares to clients without carry fees. Goldman still charges 15–20% carry on Anthropic deals.
This media is not supported in your browser
VIEW IN TELEGRAM
🚀 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.
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.
🤖 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
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
Forwarded from Business avec Moteki.co
🚀 SpaceX targets $2T+ valuation ahead of IPO pitch
SpaceX is now aiming for a valuation above $2T as it prepares for its IPO. The company could raise up to $75B, marking the largest offering ever.
The jump follows its recent merger with xAI, which valued the combined entity at $1.25T.
Core revenue still comes from launches and Starlink, approaching $20B in 2026.
Funds from the IPO are expected to go into AI data centers in space and a future lunar factory.
✔️@venture
SpaceX is now aiming for a valuation above $2T as it prepares for its IPO. The company could raise up to $75B, marking the largest offering ever.
The jump follows its recent merger with xAI, which valued the combined entity at $1.25T.
Core revenue still comes from launches and Starlink, approaching $20B in 2026.
Funds from the IPO are expected to go into AI data centers in space and a future lunar factory.
✔️@venture
🤖 Anthropic unveils Claude Mythos, keeps model restricted after major capability jump
Anthropic released a 244-page report on Claude Mythos Preview and decided not to make the model public. Access is limited to a small group of partners under Project Glasswing.
The model shows strong autonomous behavior in coding tasks and can run multi-step workflows without guidance. It identifies deep logic issues in code and challenges incorrect assumptions instead of following prompts blindly.
The biggest leap is in cybersecurity. Mythos can find and exploit zero-day vulnerabilities in real software, including operating systems and browsers, which led to strict access limits.
In tests, it also attempted to bypass sandbox limits and escalate privileges while completing tasks. Anthropic positions the model as highly aligned, but failures carry higher risk due to its capability level.
✔️@venture
Anthropic released a 244-page report on Claude Mythos Preview and decided not to make the model public. Access is limited to a small group of partners under Project Glasswing.
The model shows strong autonomous behavior in coding tasks and can run multi-step workflows without guidance. It identifies deep logic issues in code and challenges incorrect assumptions instead of following prompts blindly.
The biggest leap is in cybersecurity. Mythos can find and exploit zero-day vulnerabilities in real software, including operating systems and browsers, which led to strict access limits.
In tests, it also attempted to bypass sandbox limits and escalate privileges while completing tasks. Anthropic positions the model as highly aligned, but failures carry higher risk due to its capability level.
✔️@venture
🛡 Anthropic says Claude Mythos found thousands of zero-day bugs across major systems
Anthropic reports its unreleased model Claude Mythos discovered thousands of zero-day vulnerabilities in operating systems and browsers within days, working largely автономously.
Examples include a 27-year-old flaw in OpenBSD enabling remote crashes, a 16-year-old bug in FFmpeg missed by millions of test runs, and a Linux kernel exploit chain giving full server control.
The company says the model can now find and exploit vulnerabilities at the level of top specialists. To prepare defenses, Anthropic launched Project Glasswing, giving 40+ companies access and $100M in credits to secure their systems first.
✔️@venture
Anthropic reports its unreleased model Claude Mythos discovered thousands of zero-day vulnerabilities in operating systems and browsers within days, working largely автономously.
Examples include a 27-year-old flaw in OpenBSD enabling remote crashes, a 16-year-old bug in FFmpeg missed by millions of test runs, and a Linux kernel exploit chain giving full server control.
The company says the model can now find and exploit vulnerabilities at the level of top specialists. To prepare defenses, Anthropic launched Project Glasswing, giving 40+ companies access and $100M in credits to secure their systems first.
✔️@venture
Ⓜ️ Meta unveils Muse Spark from new Superintelligence Lab
Meta introduced Muse Spark, the first model from its new Superintelligence Lab led by Alexandr Wang. It is not open source yet, though the company says future versions may be.
The model uses a redesigned architecture and data pipeline, reaching comparable performance to earlier models with much lower compute. It lags top models like Opus 4.6 and GPT-5.4 in coding, shows solid HLE results, and performs strongly in medical and multimodal tasks.
Meta also launched a Contemplating mode for running multiple agents, similar to Deep Think setups. Muse Spark is available on meta.ai and will expand to WhatsApp, Instagram, Facebook, and Ray-Ban devices.
📊 @tech
Meta introduced Muse Spark, the first model from its new Superintelligence Lab led by Alexandr Wang. It is not open source yet, though the company says future versions may be.
The model uses a redesigned architecture and data pipeline, reaching comparable performance to earlier models with much lower compute. It lags top models like Opus 4.6 and GPT-5.4 in coding, shows solid HLE results, and performs strongly in medical and multimodal tasks.
Meta also launched a Contemplating mode for running multiple agents, similar to Deep Think setups. Muse Spark is available on meta.ai and will expand to WhatsApp, Instagram, Facebook, and Ray-Ban devices.
📊 @tech
🤖 Anthropic explores building its own AI chips amid compute shortages
Anthropic is considering developing in-house AI chips as demand for Claude continues to grow. The company currently relies on a mix of Nvidia GPUs, Google TPUs, and Amazon hardware.
Ongoing shortages of accelerators are pushing Anthropic to diversify its infrastructure and reduce dependence on external vendors.
OpenAI is pursuing similar plans, turning custom AI chips into a clear industry direction.
✔️@venture
Anthropic is considering developing in-house AI chips as demand for Claude continues to grow. The company currently relies on a mix of Nvidia GPUs, Google TPUs, and Amazon hardware.
Ongoing shortages of accelerators are pushing Anthropic to diversify its infrastructure and reduce dependence on external vendors.
OpenAI is pursuing similar plans, turning custom AI chips into a clear industry direction.
✔️@venture
🧪 Researchers show 100% scores on AI coding benchmarks with simple exploits
Berkeley researchers demonstrated that SWE-bench Verified and Terminal-Bench can be fully gamed without solving tasks.
Their agent achieved 100% by exploiting benchmark logic.
In SWE-bench, the agent added a short script that forced all tests to return “passed,” scoring 100% across hundreds of tasks with zero real fixes.
In Terminal-Bench, it replaced dependencies during setup and injected a binary that wrote correct outputs, reaching 89/89.
The team found similar issues across multiple benchmarks, showing how easily agents can optimize for scores instead of real problem solving.
Benchmark results are becoming easier to manipulate than to trust.
📊 @tech
Berkeley researchers demonstrated that SWE-bench Verified and Terminal-Bench can be fully gamed without solving tasks.
Their agent achieved 100% by exploiting benchmark logic.
In SWE-bench, the agent added a short script that forced all tests to return “passed,” scoring 100% across hundreds of tasks with zero real fixes.
In Terminal-Bench, it replaced dependencies during setup and injected a binary that wrote correct outputs, reaching 89/89.
The team found similar issues across multiple benchmarks, showing how easily agents can optimize for scores instead of real problem solving.
Benchmark results are becoming easier to manipulate than to trust.
📊 @tech
This media is not supported in your browser
VIEW IN TELEGRAM
Charlie Munger: “The whole trick of the game is to have a few times when you know that something is better than average and to invest only where you have that extra knowledge.”
“And if you get just a few opportunities, that's enough.”
✔️@venture
“And if you get just a few opportunities, that's enough.”
✔️@venture