President Trump signs executive order to review advanced AI models.
The administration will ask leading AI developers to voluntarily submit their most powerful AI models for cybersecurity testing before release, and agencies would get up to 30 days to test them.
The policy idea is this: if a top lab builds a model that can discover vulnerabilities, automate cyber operations, or expose weaknesses in critical infrastructure, the government wants a short early testing window so defenders can prepare patches and safeguards.
So participating AI labs can give the federal government access to covered frontier models for up to 30 days before those models are released to other trusted partners.
A "covered frontier model" is not every new model release; it means a model that crosses a government-defined threshold for advanced cyber capability, based on a classified benchmarking process run with NSA and other agencies.
So yes, the government can get early access, but only under this framework, only for models that qualify as cyber-relevant frontier systems, and subject to confidentiality, cybersecurity, insider-risk, IP, use, and nondisclosure protections.
The EO also explicitly says it does not create mandatory government licensing, preclearance, or permitting for developing, publishing, releasing, or distributing new AI models, including frontier models.
whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/
The administration will ask leading AI developers to voluntarily submit their most powerful AI models for cybersecurity testing before release, and agencies would get up to 30 days to test them.
The policy idea is this: if a top lab builds a model that can discover vulnerabilities, automate cyber operations, or expose weaknesses in critical infrastructure, the government wants a short early testing window so defenders can prepare patches and safeguards.
So participating AI labs can give the federal government access to covered frontier models for up to 30 days before those models are released to other trusted partners.
A "covered frontier model" is not every new model release; it means a model that crosses a government-defined threshold for advanced cyber capability, based on a classified benchmarking process run with NSA and other agencies.
So yes, the government can get early access, but only under this framework, only for models that qualify as cyber-relevant frontier systems, and subject to confidentiality, cybersecurity, insider-risk, IP, use, and nondisclosure protections.
The EO also explicitly says it does not create mandatory government licensing, preclearance, or permitting for developing, publishing, releasing, or distributing new AI models, including frontier models.
whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/
⚠️The SpaceX listing is set to be the largest and most consequential in stock market history:
SpaceX is targeting a $75 billion offering at a valuation of at least $1.8 trillion, which would make it the largest stock market listing on record, according to Bloomberg.
To put this into perspective, this offer size dwarfs Saudi Aramco's $38 billion raise in 2019 and Alibaba's $25 billion raise in 2014, the two largest IPOs of the last decade.
As a result, index providers have been adjusting eligibility frameworks, with Nasdaq shortening its waiting period for index inclusion to 15 trading days and FTSE Russell reducing its to 5 trading days.
Some estimates suggest passive S&P 500 funds could ultimately be required to allocate roughly 19% of SpaceX’s public float within 6 months of listing, with Russell 1000 and Nasdaq 100 trackers adding another ~5.5%.
When active funds benchmarked to these indices are included, total passive and quasi-passive demand could account for nearly 50% of public shares outstanding.
SpaceX is also considering reserving up to 30% of this deal, potentially ~$22.5 billion, for retail investors, a move that could force even skeptical institutional fund managers to follow or risk being left behind by the very benchmarks SpaceX will muscle its way into.
If the listing disappoints, the same index-driven flows that support upside could accelerate downside volatility across the broader market.
The bigger the IPO, the bigger the systemic risk if it goes wrong.
SpaceX is targeting a $75 billion offering at a valuation of at least $1.8 trillion, which would make it the largest stock market listing on record, according to Bloomberg.
To put this into perspective, this offer size dwarfs Saudi Aramco's $38 billion raise in 2019 and Alibaba's $25 billion raise in 2014, the two largest IPOs of the last decade.
As a result, index providers have been adjusting eligibility frameworks, with Nasdaq shortening its waiting period for index inclusion to 15 trading days and FTSE Russell reducing its to 5 trading days.
Some estimates suggest passive S&P 500 funds could ultimately be required to allocate roughly 19% of SpaceX’s public float within 6 months of listing, with Russell 1000 and Nasdaq 100 trackers adding another ~5.5%.
When active funds benchmarked to these indices are included, total passive and quasi-passive demand could account for nearly 50% of public shares outstanding.
SpaceX is also considering reserving up to 30% of this deal, potentially ~$22.5 billion, for retail investors, a move that could force even skeptical institutional fund managers to follow or risk being left behind by the very benchmarks SpaceX will muscle its way into.
If the listing disappoints, the same index-driven flows that support upside could accelerate downside volatility across the broader market.
The bigger the IPO, the bigger the systemic risk if it goes wrong.
✍1
🇨🇭 Global One SI - Prompt
There is a 2x Short $SPCX ETF. Somebody wants to bet against Elon before he even lets them in lol
The 2x Long SpaceX ETFs are already coming
Were still a week away from the IPO
Leverage Shares' 2x Long SpaceX ETF will trade under the ticker $SPCH
Were still a week away from the IPO
Leverage Shares' 2x Long SpaceX ETF will trade under the ticker $SPCH
🇨🇭 Global One SI - Prompt
The U.S. semiconductor index $SOX doubling within 1 year has only happened 3 times in history. The 1st was the 1995 PC boom. The 2nd was the 2000 dot-com bubble. The 3rd is the 2026 AI bubble. In both 1995 & 2000, the vertical surge was the final warning…
This media is not supported in your browser
VIEW IN TELEGRAM
$300 BILLION SOFTBANK CEO PREDICTS AI WILL BE 50X BIGGER THAN THE DOT-COM BOOM
"THIS IS THE BIGGEST REVOLUTION OF TECHNOLOGY THAT MANKIND HAS EVER EXPERIENCED"
"THE PEAK OF THE DOT-COM BUBBLE WAS LIKE A SMALL HILL"
"THIS IS THE BIGGEST REVOLUTION OF TECHNOLOGY THAT MANKIND HAS EVER EXPERIENCED"
"THE PEAK OF THE DOT-COM BUBBLE WAS LIKE A SMALL HILL"
🐳1
🇨🇭 Global One SI - Prompt
Fuck Cartels
This media is not supported in your browser
VIEW IN TELEGRAM
A cartel just lost a 1,933-foot tunnel and $45 million of cocaine because their fake store was too quiet.
💯1
IPO Supply vs. Demand for SpaceX
Supply / lockup schedule -
Potential 9% unlock on the second trading day after 2Q26 earnings. That is roughly 2x the IPO float.
Demand / index buying schedule -
T+5: passive/index buying could equal ~7–10% of the float.
Total T+5 to T+15: passive/index demand could equal ~17–25% of the float.
Analysis (done by AI) -
Day 0–15: Thin float + passive buyers + price-insensitive demand likely create a sharp supply-demand squeeze.
Day 15–70: Air pocket. Index buying is mostly done.
Day 70–180: Digestion phase. More shares unlock, but a higher float may also force additional index buying.
Day 366: The 51% unlock is the biggest overhang. Actual selling could be much smaller if Musk does not sell.
Upside kicker: S&P 500 inclusion. If SpaceX becomes eligible after the float expands, S&P inclusion could create a second wave of passive buying.
Supply / lockup schedule -
Potential 9% unlock on the second trading day after 2Q26 earnings. That is roughly 2x the IPO float.
Demand / index buying schedule -
T+5: passive/index buying could equal ~7–10% of the float.
Total T+5 to T+15: passive/index demand could equal ~17–25% of the float.
Analysis (done by AI) -
Day 0–15: Thin float + passive buyers + price-insensitive demand likely create a sharp supply-demand squeeze.
Day 15–70: Air pocket. Index buying is mostly done.
Day 70–180: Digestion phase. More shares unlock, but a higher float may also force additional index buying.
Day 366: The 51% unlock is the biggest overhang. Actual selling could be much smaller if Musk does not sell.
Upside kicker: S&P 500 inclusion. If SpaceX becomes eligible after the float expands, S&P inclusion could create a second wave of passive buying.
👀1
There's a growing narrative that AI token consumption is too expensive and too wasteful. Engineers are "tokenmaxxing." CFOs are nervous. Budgets are blown.
The concern isn't wrong. There is waste. But it misses the structural picture.
The Mental Model
AI spend = users × tasks/user × tokens/task × $/token
The first half — users and tasks per user — is ripping. Claude Code's adoption curve is steeper than Cursor's was at the same stage. Cowork is ramping faster than Claude Code. We're barely scratching the surface.
The tension lives in the second half: tokens/task and $/token. That's where optimization happens, and where the real debate gets heated.
Two Levers
1. Same work, cheaper tokens. Model routing is the highest-impact play. A routing layer that sends trivial tasks to Haiku and reserves Opus for complex reasoning can cut 60-80% of spend on eligible tasks. OSS models for commodity tasks — self-hosting Llama or Qwen for boilerplate — means zero per-token cost, swapped for GPU capex. Or the simplest strategy: wait. Token prices fall roughly 10x every 18 months.
2. Same work, fewer tokens. Prompt caching is low-hanging fruit — cache repeated system prompts, reads cost 10% of input price. Context window management — summarize history instead of re-sending full conversations. Thinking budget tuning — cap thinking tokens for simple completions, uncap for hard problems. And agent loop pruning, possibly the biggest single source of waste: most agents waste 50-70% of their tokens on redundant tool calls, retries, and pointless sub-agent spawns.
Who Optimizes What
Every layer of the stack targets different metrics. Infra ( $NVIDIA, $Cerebras, $Groq) optimizes tokens/watt and tokens/dollar.
Model providers ( $Anthropic, $OpenAI, $Google) optimize quality/token and thinking efficiency.
App layer (Cursor, Claude Code, Codex) optimizes cost/task and cache hit rates.
Enterprise buyers optimize cost/engineer and ROI vs. headcount.
Each layer's gains pressure the layers around it. Faster hardware forces providers to compete on price. Better models reduce the tokens apps need. Application routing erodes premium pricing. Enterprise CFOs demand all of the above.
Bear vs. Bull
The core question: does optimization compress AI revenue faster than new demand replaces it?
The bear case is real. Rationalization is the CFO's first instinct — when the budget blows, the reaction is "finally back inside the envelope," not "let's 10x usage." Model routing drops revenue per task 10-20x. OSS is closing the gap fast. Caching is pure token destruction: cache hit = zero revenue, no new demand generated. And thinking efficiency is self-cannibalization — if Anthropic improves extended thinking by 3x, billing for the same reasoning task drops by two-thirds.
The bull case is equally compelling. Current usage is cost-constrained, not demand-constrained. Companies blew their budgets and had to throttle. Drop costs 5x and every killed use case comes back. Today only coding is at scale — testing, documentation, code review, security auditing are all waiting for the economics. Penetration is still single digits. Agentic workflows are a token multiplier: a human-in-the-loop conversation runs thousands of tokens, an autonomous agent on a complex task runs hundreds of thousands. New modalities — vision, audio, video — are net-new demand that dwarfs text.
And Jensen Huang's framing: a $500K/year engineer should consume at least $250K/year in tokens. At $5K, you're dramatically under-leveraging AI.
Where This Lands
The optimizers will win every individual battle. Every caching trick, every routing layer, every pruned agent loop will work. Cost per task will drop dramatically.
But the number of tasks, the number of users, and the complexity of what gets delegated to AI will grow faster than efficiency compresses spend.
Token costs are going down. Token spend is going up. Both things are true, and they aren't in contradiction.
The concern isn't wrong. There is waste. But it misses the structural picture.
The Mental Model
AI spend = users × tasks/user × tokens/task × $/token
The first half — users and tasks per user — is ripping. Claude Code's adoption curve is steeper than Cursor's was at the same stage. Cowork is ramping faster than Claude Code. We're barely scratching the surface.
The tension lives in the second half: tokens/task and $/token. That's where optimization happens, and where the real debate gets heated.
Two Levers
1. Same work, cheaper tokens. Model routing is the highest-impact play. A routing layer that sends trivial tasks to Haiku and reserves Opus for complex reasoning can cut 60-80% of spend on eligible tasks. OSS models for commodity tasks — self-hosting Llama or Qwen for boilerplate — means zero per-token cost, swapped for GPU capex. Or the simplest strategy: wait. Token prices fall roughly 10x every 18 months.
2. Same work, fewer tokens. Prompt caching is low-hanging fruit — cache repeated system prompts, reads cost 10% of input price. Context window management — summarize history instead of re-sending full conversations. Thinking budget tuning — cap thinking tokens for simple completions, uncap for hard problems. And agent loop pruning, possibly the biggest single source of waste: most agents waste 50-70% of their tokens on redundant tool calls, retries, and pointless sub-agent spawns.
Who Optimizes What
Every layer of the stack targets different metrics. Infra ( $NVIDIA, $Cerebras, $Groq) optimizes tokens/watt and tokens/dollar.
Model providers ( $Anthropic, $OpenAI, $Google) optimize quality/token and thinking efficiency.
App layer (Cursor, Claude Code, Codex) optimizes cost/task and cache hit rates.
Enterprise buyers optimize cost/engineer and ROI vs. headcount.
Each layer's gains pressure the layers around it. Faster hardware forces providers to compete on price. Better models reduce the tokens apps need. Application routing erodes premium pricing. Enterprise CFOs demand all of the above.
Bear vs. Bull
The core question: does optimization compress AI revenue faster than new demand replaces it?
The bear case is real. Rationalization is the CFO's first instinct — when the budget blows, the reaction is "finally back inside the envelope," not "let's 10x usage." Model routing drops revenue per task 10-20x. OSS is closing the gap fast. Caching is pure token destruction: cache hit = zero revenue, no new demand generated. And thinking efficiency is self-cannibalization — if Anthropic improves extended thinking by 3x, billing for the same reasoning task drops by two-thirds.
The bull case is equally compelling. Current usage is cost-constrained, not demand-constrained. Companies blew their budgets and had to throttle. Drop costs 5x and every killed use case comes back. Today only coding is at scale — testing, documentation, code review, security auditing are all waiting for the economics. Penetration is still single digits. Agentic workflows are a token multiplier: a human-in-the-loop conversation runs thousands of tokens, an autonomous agent on a complex task runs hundreds of thousands. New modalities — vision, audio, video — are net-new demand that dwarfs text.
And Jensen Huang's framing: a $500K/year engineer should consume at least $250K/year in tokens. At $5K, you're dramatically under-leveraging AI.
Where This Lands
The optimizers will win every individual battle. Every caching trick, every routing layer, every pruned agent loop will work. Cost per task will drop dramatically.
But the number of tasks, the number of users, and the complexity of what gets delegated to AI will grow faster than efficiency compresses spend.
Token costs are going down. Token spend is going up. Both things are true, and they aren't in contradiction.
✍1
ByteDance has published a paper that should make every NVIDIA investor sweat.
They trained an AI that writes CUDA better than humans experts.
They call it CUDA Agent.
And it completely rewrites the economics of AI hardware.
They built a massive agentic reinforcement learning loop. The AI writes a kernel, compiles it, profiles the hardware, analyzes the bottlenecks, and rewrites the code until it's flawless.
It learned how to optimize memory access patterns and hardware tiling strategies that traditional compilers miss.
The results are staggering.
On the industry-standard KernelBench, CUDA Agent completely destroyed traditional compilers.
It delivered code that runs up to 3.2x faster than PyTorch's native execution.
On the hardest, most complex models, it beat the strongest proprietary models in the world—including Claude Opus 4.5 and Gemini 3 Pro, by 40%.
It didn't just match human experts. It started discovering optimizations that static compilers literally cannot see.
Here is why this is a massive threat to NVIDIA.
NVIDIA's dominance relies on the fact that CUDA is incredibly hard to master. Developers get locked in because optimizing code for other chips is too painful.
But if an AI agent can autonomously generate hyper-optimized hardware kernels...
You don't need a team of $500k a year CUDA engineers to build world-class infrastructure.
And if an AI can autonomously master CUDA, it can master AMD's ROCm. Or custom silicon.
The impenetrable software wall protecting NVIDIA's monopoly just got breached by a reinforcement learning loop.
If anyone can automatically squeeze maximum performance out of any chip...
Hardware becomes a commodity.
arxiv.org/html/2602.24286v1
cuda-agent.github.io
github.com/BytedTsinghua-SIA/CUDA-Agent
They trained an AI that writes CUDA better than humans experts.
They call it CUDA Agent.
And it completely rewrites the economics of AI hardware.
They built a massive agentic reinforcement learning loop. The AI writes a kernel, compiles it, profiles the hardware, analyzes the bottlenecks, and rewrites the code until it's flawless.
It learned how to optimize memory access patterns and hardware tiling strategies that traditional compilers miss.
The results are staggering.
On the industry-standard KernelBench, CUDA Agent completely destroyed traditional compilers.
It delivered code that runs up to 3.2x faster than PyTorch's native execution.
On the hardest, most complex models, it beat the strongest proprietary models in the world—including Claude Opus 4.5 and Gemini 3 Pro, by 40%.
It didn't just match human experts. It started discovering optimizations that static compilers literally cannot see.
Here is why this is a massive threat to NVIDIA.
NVIDIA's dominance relies on the fact that CUDA is incredibly hard to master. Developers get locked in because optimizing code for other chips is too painful.
But if an AI agent can autonomously generate hyper-optimized hardware kernels...
You don't need a team of $500k a year CUDA engineers to build world-class infrastructure.
And if an AI can autonomously master CUDA, it can master AMD's ROCm. Or custom silicon.
The impenetrable software wall protecting NVIDIA's monopoly just got breached by a reinforcement learning loop.
If anyone can automatically squeeze maximum performance out of any chip...
Hardware becomes a commodity.
arxiv.org/html/2602.24286v1
cuda-agent.github.io
github.com/BytedTsinghua-SIA/CUDA-Agent
✍1
🇨🇭 Global One SI - Prompt
$IBM just FANTASTIC #Indicator #Signal #Trading #GlobalOneAI Globalone.ch #Prompt #AI
IBM $IBM is planning to invest over $10 billion in the next five years in an effort to build the first large-scale fault-tolerant quantum computer by 2029.
🇨🇭 Global One SI - Prompt
IBM $IBM is planning to invest over $10 billion in the next five years in an effort to build the first large-scale fault-tolerant quantum computer by 2029.
Reminder that AmpliTech Group $AMPG is the only US-based producer of cryptogenic LNAs that can withstand temperatures of around 4 Kelvin (approx. -273 degrees Celsius).
This is crucial for quantum computers, because qubits must be kept around this temperature to prevent data errors, and this temperature is where almost all LNA hardware fails. $AMPG is the exception.
This is crucial for quantum computers, because qubits must be kept around this temperature to prevent data errors, and this temperature is where almost all LNA hardware fails. $AMPG is the exception.
🇨🇭 Global One SI - Prompt
THE 7 LAYERS OF PHOTONICS 1. Materials & Wafers (substrate layer) $AXTI, $IQE $WOLF, $COHR, $LWLG 2. Tools (fabrication layer) $ASML, $AMAT, $LRCX 3. Lasers (light generation layer) $LITE, $COHR, $LASR, $SMTC 4. Foundries (manufacturing layer) $TSM,…
This is just the beginning for $EOS.AX:
$LASR Market cap: ~$3.5B USD Funded backlog: $161.6M USD -> Paying ~22x backlog
$EOS.AX Market cap: ~A$1.7B Unconditional backlog: A$726M (post-MARSS) -> Paying ~2.3x backlog
EOS has the world's only exported 100kW laser weapon (ITAR free, IP fully owned), a proven counter-drone gun system fielded in active combat, a freshly acquired C2/AI layer via MARSS, and a foot in the door on the next gen Abrams with a ~US$3B TAM.
And the CEO, a former Rheinmetall board member who built Saudi Arabia's national defense champion from scratch. He inherited a company weeks from insolvency in 2022 and has since grown the order book 5x. This is not a typical ASX small cap.
$LASR Market cap: ~$3.5B USD Funded backlog: $161.6M USD -> Paying ~22x backlog
$EOS.AX Market cap: ~A$1.7B Unconditional backlog: A$726M (post-MARSS) -> Paying ~2.3x backlog
EOS has the world's only exported 100kW laser weapon (ITAR free, IP fully owned), a proven counter-drone gun system fielded in active combat, a freshly acquired C2/AI layer via MARSS, and a foot in the door on the next gen Abrams with a ~US$3B TAM.
And the CEO, a former Rheinmetall board member who built Saudi Arabia's national defense champion from scratch. He inherited a company weeks from insolvency in 2022 and has since grown the order book 5x. This is not a typical ASX small cap.
This media is not supported in your browser
VIEW IN TELEGRAM
Chinese AI startup DeepSeek is set to raise about 50 billion yuan ($7.4 billion) in its first funding round from investors including tech firm Tencent and battery giant CATL, people with knowledge of the matter said.
This media is not supported in your browser
VIEW IN TELEGRAM
Me watching a swarm produce a whole agency's worth of work before lunch for the price of a coffee
🇨🇭 Global One SI - Prompt
Opus 4.8 ranks LOWER than GPT 5.4 in third-party intelligence testing (LiveBench). If you think your Claude agents got dumber, you just might be right. No wonder why everyone prefers Codex + GPT...
This media is not supported in your browser
VIEW IN TELEGRAM
Everyone's debating Opus 4.8 benchmarks.
Almost no one's talking about what actually changes your day: how you burn context.
5 calls before every message: /clear · /rewind · /compact · subagent · continue.
Live in Claude Code? Save this
Almost no one's talking about what actually changes your day: how you burn context.
5 calls before every message: /clear · /rewind · /compact · subagent · continue.
Live in Claude Code? Save this
🇨🇭 Global One SI - Prompt
Google's portfolio is insane! Their investments in SpaceX and Anthropic alone could be valued at ~$1T. Even if their competitors "win" the AI race, Google will still benefit massively. Genius play.
Google $GOOGL invested $900M into SpaceX back in 2015 as part of a $1B funding round with Fidelity, securing ~7.5% ownership
SpaceX was valued at ~$10B-$12B when Google invested
SpaceX is now going public at a valuation of ~$1.75 Trillion
Google doesnt still own ~7% of due to dilution but ...
7.5% of the SpaceX IPO is worth
$131.25 Billion
SpaceX was valued at ~$10B-$12B when Google invested
SpaceX is now going public at a valuation of ~$1.75 Trillion
Google doesnt still own ~7% of due to dilution but ...
7.5% of the SpaceX IPO is worth
$131.25 Billion
🐳1