This media is not supported in your browser
VIEW IN TELEGRAM
🔥🇨🇳China showcases seamless teamwork: man, drone, and robot in enemy clearance.
This media is not supported in your browser
VIEW IN TELEGRAM
🚨🇨🇳 China flies massive cargo drone
The Changying-8 hauls 3.5 tons and can take off at 7 tons. Fully loaded, it can reach 3,000 kilometers.
The Changying-8 hauls 3.5 tons and can take off at 7 tons. Fully loaded, it can reach 3,000 kilometers.
This media is not supported in your browser
VIEW IN TELEGRAM
Secretary of State Marco Rubio explains why the U.S. decided to strike Iran
It’s under two minutes and worth a watch
It’s under two minutes and worth a watch
👀1
This media is not supported in your browser
VIEW IN TELEGRAM
🇺🇸 Donald Trump Mocks Iran:
Reporter: "Iran Said they Could Target US Companies"
Trump: “With What, BB Guns? They’re Don't Even Have Nukes”.
Reporter: "Iran Said they Could Target US Companies"
Trump: “With What, BB Guns? They’re Don't Even Have Nukes”.
🇨🇭 Global One SI - Prompt
🚨 THE EU JUST ISSUED ENERGY LOCKDOWNS AND NOBODY IS CALLING IT WHAT IT IS The European Commission is urging Europeans to work from home, drive less, and fly less because of the Gulf conflict. EU Energy Commissioner Dan Jørgensen says this crisis won't end…
🇩🇪 Germany had it figured out in 1974
Then they changed their minds.
This chart is from the German Economics Ministry's own energy model... 52 years ago.
The plan was clear:
☢️ Nuclear takes over
🛢️ Oil peaks and fades
⚡ A clean, stable energy future
Germany was on track.
Then they shut it all down.
Pressured by politics.
The last German nuclear plant closed in 2023.
Now Germany is begging for cheap and reliable energy
The engineers in 1974 got the future exactly right.
The politicians in 2023 chose a different path.
And now ordinary Germans are paying the price.
One of the the most expensive energy mistake in modern history wasn't made by accident.
It was voted for.
Then they changed their minds.
This chart is from the German Economics Ministry's own energy model... 52 years ago.
The plan was clear:
☢️ Nuclear takes over
🛢️ Oil peaks and fades
⚡ A clean, stable energy future
Germany was on track.
Then they shut it all down.
Pressured by politics.
The last German nuclear plant closed in 2023.
Now Germany is begging for cheap and reliable energy
The engineers in 1974 got the future exactly right.
The politicians in 2023 chose a different path.
And now ordinary Germans are paying the price.
One of the the most expensive energy mistake in modern history wasn't made by accident.
It was voted for.
✍1
This media is not supported in your browser
VIEW IN TELEGRAM
🇦🇺The Australian Prime Minister is begging citizens to take the bus🚍
This is what an energy crisis looks like in a G20 country.
Not a press release.
Not a policy paper.
A head of government on live TV asking ordinary people to stop using fuel.
The managed demand destruction has officially begun.
There are 2 versions of demand destruction:
📋 The managed one governments ask nicely
💥 The unmanaged one prices rise until economies break
Australia just entered phase 1
The question is how long before "please take the train" becomes "you have no choice"?
This is what an energy crisis looks like in a G20 country.
Not a press release.
Not a policy paper.
A head of government on live TV asking ordinary people to stop using fuel.
The managed demand destruction has officially begun.
There are 2 versions of demand destruction:
📋 The managed one governments ask nicely
💥 The unmanaged one prices rise until economies break
Australia just entered phase 1
The question is how long before "please take the train" becomes "you have no choice"?
✍1
This media is not supported in your browser
VIEW IN TELEGRAM
Trump just said it out loud.
The US is almost done.
"If France wants oil they can fend for themselves."
For 80 years, the US Navy kept the world's most critical energy chokepoint open.
That era just ended.
20 million barrels/day flow through Hormuz.
Europe depends on it.
Asia depends on it.
The entire global oil market depends on it.
And Washington just said: not our problem.
Every barrel that moves through that strait now carries a new risk premium.
Forever?
The US is almost done.
"If France wants oil they can fend for themselves."
For 80 years, the US Navy kept the world's most critical energy chokepoint open.
That era just ended.
20 million barrels/day flow through Hormuz.
Europe depends on it.
Asia depends on it.
The entire global oil market depends on it.
And Washington just said: not our problem.
Every barrel that moves through that strait now carries a new risk premium.
Forever?
👀1
🇨🇭 Global One SI - Prompt
After getting setup with an OpenClaw instance, everyone is wondering “what’s next”? What should I build? And now we have a hassle-free way to use OpenClaw KiloClaw is managed OpenClaw. You can provision an OpenClaw agent from the Kilo Dashboard, and use…
This media is not supported in your browser
VIEW IN TELEGRAM
Introducing Skillful.sh!
It crawls and indexes 55+ directories, registries, and package managers across the AI agent ecosystem.
Right now, it's tracking 314,000+ tools across MCP servers, AI skills, and AI agents.
Every tool gets a security score (A+ through F) based on static analysis, dependency vulnerabilities, and AI code review.
You can filter by category, security grade, star count, downloads, license, whatever you need.
One search instead of twelve tabs. It's free to use, and as a developer, you can publish your package to all directories at once!
No need to worry about the exact formatting requirements for publishing to each service!
skillful.sh
It crawls and indexes 55+ directories, registries, and package managers across the AI agent ecosystem.
Right now, it's tracking 314,000+ tools across MCP servers, AI skills, and AI agents.
Every tool gets a security score (A+ through F) based on static analysis, dependency vulnerabilities, and AI code review.
You can filter by category, security grade, star count, downloads, license, whatever you need.
One search instead of twelve tabs. It's free to use, and as a developer, you can publish your package to all directories at once!
No need to worry about the exact formatting requirements for publishing to each service!
skillful.sh
✍1
This media is not supported in your browser
VIEW IN TELEGRAM
🇩🇪 BREAKING: Germany is moving to mass-produce the Skynex air defense system and Skyranger air defense system gun-based air defense systems.
Shahed-style drone attacks are pushing Europe toward the short-range anti-drone layer already used in combat in Ukraine, according to reports from UNITED24.
Shahed-style drone attacks are pushing Europe toward the short-range anti-drone layer already used in combat in Ukraine, according to reports from UNITED24.
This media is not supported in your browser
VIEW IN TELEGRAM
Optimus v3 didn’t arrive as scheduled.
But someone made a video imagining a “household butler” version of Optimus:
Cooking, folding laundry, carrying shopping bags, dancing…
Feels more like a new member of the family.
Is this what we want?
But someone made a video imagining a “household butler” version of Optimus:
Cooking, folding laundry, carrying shopping bags, dancing…
Feels more like a new member of the family.
Is this what we want?
The market is completely sleeping on the real alpha from Nvidia’s latest $2B optical plays $COHR, $LITE and $MRVL 🔥
This isn’t just another round of hyperscaler optics hype. It’s Nvidia repeating the exact same 2024 playbook, but this time for the CPO / 1.6T+ / Optical I/O architecture shift. Last year they pre-allocated EML capacity and we watched Lumentum explode +955%.
Now they’re doing it upstream again. The bottlenecks aren’t in the finished transceivers anymore. They’re in the CW/ELS lasers, wafer-level reliability, and photonics-grade substrates.
$SIVE - The CW Laser Chokepoint Designed Into Marvell’s Entire CPO Roadmap
Sivers Photonics isn’t some random laser shop. They are the confirmed light-source partner for Jabil’s reference designs and Marvell’s newly acquired Celestial AI Photonic Fabric platform.
Sivers high-power InP laser arrays are already qualified and designed-in for Celestial’s optical I/O engines - the exact tech Marvell just paid up to $5.5B to own. Marvell themselves guided Celestial to $500M ARR by end of FY28 and $1B by FY29.
And here’s the alpha most analysts still don’t have on their models: Sivers is one of the only suppliers that can feed both major CPO camps right now (Marvell/Celestial + Ayar Labs ecosystem). Dual-sourced into hyperscaler reference designs.
$AEHR - The Gatekeeper That Just Got a Fresh Hyperscale Validation
Yesterday Aehr announced a major new silicon-photonics customer - a global networking leader and key data-center optical supplier, placing an initial order for high-power FOX-XP wafer-level burn-in systems configured for nine wafers in parallel.
Aehr is the undisputed leader in multi-wafer, high-power (up to 3,500W/wafer) WLBI for silicon photonics ICs. They’re the only ones who have already shipped production-capable multi-wafer solutions. In an industry where infant mortality and long-term reliability are non-negotiable for hyperscale AI clusters running 24/7, you cannot ship at scale without Aehr’s platform.
This new order is direct proof that the Marvell/Nvidia optical supply-chain push is now moving from qualification → full commercial ramp.
$SOI - The Invisible Monopoly
Every single advanced silicon-photonics engine from Nvidia, Broadcom, and Marvell sits on Photonics-SOI substrates.
And there is effectively one company on Earth that can supply the high-volume, photonics-grade SOI wafers the foundries will actually accept: Soitec.
Their Smart Cut technology gives them a two-decade moat in precise 220nm/340nm device layers required for low-loss waveguides and modulators. Regular SOI won’t cut it for CPO - you need the photonics-tuned version. Capacity is brutally tight and scaling it is far harder than people think.
Every time Marvell or Nvidia signs another multibillion optics deal, $SOI’s order book quietly gets fatter.
This isn’t just another round of hyperscaler optics hype. It’s Nvidia repeating the exact same 2024 playbook, but this time for the CPO / 1.6T+ / Optical I/O architecture shift. Last year they pre-allocated EML capacity and we watched Lumentum explode +955%.
Now they’re doing it upstream again. The bottlenecks aren’t in the finished transceivers anymore. They’re in the CW/ELS lasers, wafer-level reliability, and photonics-grade substrates.
$SIVE - The CW Laser Chokepoint Designed Into Marvell’s Entire CPO Roadmap
Sivers Photonics isn’t some random laser shop. They are the confirmed light-source partner for Jabil’s reference designs and Marvell’s newly acquired Celestial AI Photonic Fabric platform.
Sivers high-power InP laser arrays are already qualified and designed-in for Celestial’s optical I/O engines - the exact tech Marvell just paid up to $5.5B to own. Marvell themselves guided Celestial to $500M ARR by end of FY28 and $1B by FY29.
And here’s the alpha most analysts still don’t have on their models: Sivers is one of the only suppliers that can feed both major CPO camps right now (Marvell/Celestial + Ayar Labs ecosystem). Dual-sourced into hyperscaler reference designs.
$AEHR - The Gatekeeper That Just Got a Fresh Hyperscale Validation
Yesterday Aehr announced a major new silicon-photonics customer - a global networking leader and key data-center optical supplier, placing an initial order for high-power FOX-XP wafer-level burn-in systems configured for nine wafers in parallel.
Aehr is the undisputed leader in multi-wafer, high-power (up to 3,500W/wafer) WLBI for silicon photonics ICs. They’re the only ones who have already shipped production-capable multi-wafer solutions. In an industry where infant mortality and long-term reliability are non-negotiable for hyperscale AI clusters running 24/7, you cannot ship at scale without Aehr’s platform.
This new order is direct proof that the Marvell/Nvidia optical supply-chain push is now moving from qualification → full commercial ramp.
$SOI - The Invisible Monopoly
Every single advanced silicon-photonics engine from Nvidia, Broadcom, and Marvell sits on Photonics-SOI substrates.
And there is effectively one company on Earth that can supply the high-volume, photonics-grade SOI wafers the foundries will actually accept: Soitec.
Their Smart Cut technology gives them a two-decade moat in precise 220nm/340nm device layers required for low-loss waveguides and modulators. Regular SOI won’t cut it for CPO - you need the photonics-tuned version. Capacity is brutally tight and scaling it is far harder than people think.
Every time Marvell or Nvidia signs another multibillion optics deal, $SOI’s order book quietly gets fatter.
✍1
New MIT, Blackrock, JP Morgan and other top labs paper shows AI is able to catch when management changes the metric it uses to present performance.
Companies quietly change which metrics they emphasize on earnings calls, and that shift often signals underperformance.
That pattern is called “moving targets.” When metric turnover is high in 1 quarter, weak stock performance often follows in the months that come next.
The problem is that older text methods miss this because they rely on exact wording and often mistake vague fragments for real metrics.
The signal is not simply that companies use different words over time.
It is that they may stop highlighting the metrics that used to define success, and that shift can tell investors something before the stock fully reflects it.
Older keyword systems struggle because language in earnings calls is messy.
They collapse “North America cloud revenue” into “revenue,” fail to link “sales growth” with “top-line expansion,” and often extract noise that is not really a performance metric at all.
The LLM approach does something more sensible.
It first pulls out full metric phrases with their qualifiers, then uses embeddings to judge whether this quarter’s language still matches the metrics emphasized in the same quarter a year earlier.
When that richer measure is used, firms with high “moving targets” underperform firms with low ones, with a risk-adjusted long-short spread of roughly half a percent a month.
The older NER baseline mostly surfaces junk phrases like “the %” and “the range,” and in the regression tests it even leans in the wrong direction rather than cleanly capturing weakness.
This does not prove that changing the subject causes underperformance.
It shows something more useful: if you measure managerial emphasis with enough semantic precision, what companies stop talking about becomes financially informative.
They test this on S&P 100 earnings calls from 2010 to 2024, covering 5,615 firm-quarter observations, and compare the method with a named entity recognition baseline, which is a common tool for tagging words in text.
The LLM system pulls out real business measures like market share, cash flow, and revenue growth, and its high-shift group underperforms the low-shift group with a significant 5-factor alpha of negative 0.52%, while the older method is weak and even points the wrong way in regressions.
----
Paper Link – arxiv.org/abs/2510.03195
Paper Title: "From Text to Alpha: Can LLMs Track Evolving Signals in Corporate Disclosures?"
Companies quietly change which metrics they emphasize on earnings calls, and that shift often signals underperformance.
That pattern is called “moving targets.” When metric turnover is high in 1 quarter, weak stock performance often follows in the months that come next.
The problem is that older text methods miss this because they rely on exact wording and often mistake vague fragments for real metrics.
The signal is not simply that companies use different words over time.
It is that they may stop highlighting the metrics that used to define success, and that shift can tell investors something before the stock fully reflects it.
Older keyword systems struggle because language in earnings calls is messy.
They collapse “North America cloud revenue” into “revenue,” fail to link “sales growth” with “top-line expansion,” and often extract noise that is not really a performance metric at all.
The LLM approach does something more sensible.
It first pulls out full metric phrases with their qualifiers, then uses embeddings to judge whether this quarter’s language still matches the metrics emphasized in the same quarter a year earlier.
When that richer measure is used, firms with high “moving targets” underperform firms with low ones, with a risk-adjusted long-short spread of roughly half a percent a month.
The older NER baseline mostly surfaces junk phrases like “the %” and “the range,” and in the regression tests it even leans in the wrong direction rather than cleanly capturing weakness.
This does not prove that changing the subject causes underperformance.
It shows something more useful: if you measure managerial emphasis with enough semantic precision, what companies stop talking about becomes financially informative.
They test this on S&P 100 earnings calls from 2010 to 2024, covering 5,615 firm-quarter observations, and compare the method with a named entity recognition baseline, which is a common tool for tagging words in text.
The LLM system pulls out real business measures like market share, cash flow, and revenue growth, and its high-shift group underperforms the low-shift group with a significant 5-factor alpha of negative 0.52%, while the older method is weak and even points the wrong way in regressions.
----
Paper Link – arxiv.org/abs/2510.03195
Paper Title: "From Text to Alpha: Can LLMs Track Evolving Signals in Corporate Disclosures?"
✍1
🇨🇭 Global One SI - Prompt
$ORCL/USD H2 #Indicator #Signal #Trading #GlobalOneAI
🚨 NEWS: $ORCL Oracle just began one of its biggest restructurings in years, cutting jobs across the US, India, and other offices with same-day termination emails as it redirects money toward AI data centers.
Employees receive email from 'Oracle Leadership' at 6AM, saying: "We have made the decision to eliminate your role as…"
The reported logic is brutally financial: Oracle appears to be trimming operating costs so it can fund a capital-heavy buildout of the servers, chips, networking gear, and power systems needed to rent AI computing at large scale.
Reports say affected teams include RHS, SVOS, and NetSuite IDC, with some units describing cuts near 30%, which suggests Oracle is not making small corrections but reshaping multiple parts of the company around infrastructure priorities.
The striking part is that this is happening alongside strong recent profits, which shows Oracle is treating the AI buildout as urgent enough to favor future capacity over current headcount stability.
Employees receive email from 'Oracle Leadership' at 6AM, saying: "We have made the decision to eliminate your role as…"
The reported logic is brutally financial: Oracle appears to be trimming operating costs so it can fund a capital-heavy buildout of the servers, chips, networking gear, and power systems needed to rent AI computing at large scale.
Reports say affected teams include RHS, SVOS, and NetSuite IDC, with some units describing cuts near 30%, which suggests Oracle is not making small corrections but reshaping multiple parts of the company around infrastructure priorities.
The striking part is that this is happening alongside strong recent profits, which shows Oracle is treating the AI buildout as urgent enough to favor future capacity over current headcount stability.
👀1
🇨🇭 Global One SI - Prompt
Claude code source code has been leaked via a map file in their npm registry! Code: github.com/instructkr/claude-code Backed the source up on github
So yesterday's massive Anthropic source-code leak was a process bug, a “human error”.
Usually means the system still depended on memory, attention, and luck at exactly the point where it should have depended on automated-guardrails.
His most interesting line is the counterintuitive one: solve it by going faster, not by adding more process.
Slow, manual workflows create more opportunities for drift, more handoffs, more stale assumptions, and more moments where one tired person becomes the last line of defense.
Fast systems are safer when speed comes from automation.
A good deploy pipeline is not just efficient. It is hostile to improvisation. It makes the right action the default, makes the wrong action hard, and checks the artifact that actually ships rather than the one everyone thinks is shipping.
Usually means the system still depended on memory, attention, and luck at exactly the point where it should have depended on automated-guardrails.
His most interesting line is the counterintuitive one: solve it by going faster, not by adding more process.
Slow, manual workflows create more opportunities for drift, more handoffs, more stale assumptions, and more moments where one tired person becomes the last line of defense.
Fast systems are safer when speed comes from automation.
A good deploy pipeline is not just efficient. It is hostile to improvisation. It makes the right action the default, makes the wrong action hard, and checks the artifact that actually ships rather than the one everyone thinks is shipping.
✍1