Glimpse.
All-In Podcast
1 hr 42 min video
3 min read
AI Sovereignty, Anthropic's Betrayal, SCOTUS Birthright Ruling, California's Fiscal Collapse
You just saved 1 hr 39 min.
The big takeaway
Palantir and Nvidia partner on sovereign AI to keep government data domestic, exposing how frontier labs like Anthropic exploit enterprise data to build competing products. SCOTUS upholds birthright citizenship 6-3. California's $351B budget masks $20-40B in hidden debt, massive tax exodus, and $1.5-2T in unfunded liabilities, threatening state solvency.
AI Sovereignty and the Frontier Lab Trap
Palantir-Nvidia Sovereign AI Partnership
Palantir and Nvidia announced a partnership to build a custom frontier-quality AI model for US government agencies, where the government owns the hardware, data, and model weights. This addresses concerns that proprietary data shared with frontier labs gets exploited to build competing products.
Anthropic's Vertical Integration Playbook
Anthropic has systematically launched competing products in categories where customers built successful applications on top of Claude: Claude Design (competing with Figma), Claude Code (competing with Cursor), Claude Security, Claude Legal, Claude Financial. This mirrors Microsoft's and Google's historical playbook of using platform dominance to capture lucrative verticals.
1
Claude Design
Competed with Figma
2
Claude Code
Competed with Cursor
3
Claude Security
New vertical
4
Claude Legal
New vertical
5
Claude Financial
New vertical
Anthropic's vertical integration into customer categories
Figma's Cautionary Tale
Anthropic's chief product officer sat on Figma's board and did not resign until 3 days before Anthropic launched Claude Design, a direct competitor. Figma's stock fell 50% in the year following, illustrating the risk enterprises face when sharing data with frontier labs.
Before Claude Design Launch
Figma Board Member (CPO)
After Claude Design Launch
Figma Stock Down 50%
Figma's experience with Anthropic's competitive move
Enterprise Cost and Performance: Open Source vs. Claude
8090's testing showed that wrapping Claude with their software factory harness achieved 1.4x cheaper and 1.5x faster results than Claude alone. Using an open-source frontier model with their harness was 16.4x cheaper, though 3x slower—a trade-off many enterprises find acceptable for massive cost savings and data sovereignty.
Claude alone
1 baseline cost
8090 + Claude
0.71 cost (1.4x cheaper)
8090 + Open Source
0.061 cost (16.4x cheaper)
Cost comparison: frontier vs. open-source models with harness
Intelligence Sovereignty vs. Privacy
Intelligence sovereignty means enterprises retain control over which AI analyzes their data and how it interprets their world—distinct from privacy, which is about preventing data access. It ensures competitors cannot use shared data to build better products against you.
The Distributed Inference Future
Enterprises are shifting from a hub-and-spoke cloud model to a distributed model: large hubs for foundational training, medium hubs for enterprise fine-tuning on proprietary data, and distributed spokes for local inference. This allows companies to run workflows on-premises, avoiding cloud dependency and data leakage.
1
Large hubs: foundational model training
2
Medium hubs: enterprise fine-tuning on proprietary data
3
Distributed spokes: local on-premises inference
4
Result: data stays internal, costs drop, control retained
Evolution from centralized to distributed AI infrastructure
Nvidia's Open-Source Strategy Shift
Nvidia has historically downplayed its open-source Neotron model to avoid alarming major customers. After OpenAI, Anthropic, and AMD announced competing chips and Elon signaled his own fab, Nvidia took the gloves off and began promoting Neotron as a competitive alternative to closed models, positioning itself as a full-stack provider.
Why Chip and App Companies Want Model Competition
Chip makers (Nvidia), applications (Palantir), and enterprises all benefit from a competitive model layer because it prevents lock-in and monopsony pricing. Only Anthropic and OpenAI want duopoly dominance. A healthy competitive model layer ensures diverse buyers for chips and prevents any single model provider from controlling enterprise destiny.
Token Price Deflation Ahead
With open-source competition intensifying, token costs are projected to fall 90% annually for the next 3 years, making enterprise AI dramatically cheaper and enabling companies to run wasteful experiments locally without concern.
90%
Projected annual token price decline over 3 years
Deflationary impact of open-source competition
Anthropic Export Controls and Palace Intrigue
Mythos Jailbreak and Export Restrictions
Anthropic's Mythos model was found to have a jailbreak that bypassed safety guardrails when tested by Amazon. Dario Amodei had previously called Mythos a cyber weapon. The US Commerce Department issued export restrictions on Fable and Mythos on June 30, but lifted them on June 26 after Anthropic replaced Dario with co-founder Tom Brown as lead negotiator.
June 26
Export restrictions lifted by Commerce Dept
June 30
Export restrictions originally issued
Post-June 26
Tom Brown replaces Dario as lead negotiator
Timeline of Anthropic export control events
Three Conditions for Export Control Letter
The export control letter resulted from three specific conditions: (1) Dario publicly calling Mythos a cyber weapon, (2) Amazon reporting the jailbreak failure to Commerce, and (3) Dario refusing to roll back Fable until the jailbreak was fixed. Change any one condition and the letter would not have been issued.
1
Dario calls Mythos a cyber weapon
2
Amazon reports jailbreak failure to Commerce
3
Dario refuses to roll back Fable
4
Result: Export restrictions issued
Three conditions that triggered export controls
Open-Source Models Are Not Chinese Once Forked
Once a model like DeepSeek or Qwen is open-sourced, it ceases to be inherently Chinese because enterprises can fork it, run it on US hardware in US data centers with no data leakage to China. Banning open-source imports would put the US at a disadvantage globally and subject American companies to a token tax.
SCOTUS Birthright Citizenship Ruling
Supreme Court Upholds Birthright Citizenship 6-3
Chief Justice Roberts, joined by the three liberal justices, ruled that the 14th Amendment guarantees automatic citizenship to children born in the US regardless of parental immigration status. Trump's executive order ending this was struck down. Justice Kavanaugh suggested Congress could still tighten birthright rules consistent with the Constitution.
255,000
Children born annually to non-citizen parents in US
Population affected by birthright citizenship ruling
Original Intent vs. Textual Reading Debate
The 14th Amendment was ratified in 1868 to ensure freed slaves and their descendants had citizenship rights after Dred Scott denied it. Intentionalists argue the framers never contemplated modern immigration; textualists read the plain language as granting citizenship to all born in the US subject to its jurisdiction.
Proposed Compromise: Residency-Based Citizenship
A middle-ground proposal would grant birthright citizenship only to children of legal residents (green card holders), not temporary visitors or illegal immigrants. This preserves citizenship for long-term residents while excluding birth tourism and casual visitors.
Birth Tourism Example: Dong Wan Lee Case
In 2019, Dong Wan Lee pleaded guilty to helping pregnant Chinese nationals enter the US under false pretenses so their babies would gain US citizenship. She served 10 months in prison. This case illustrates the birth tourism concern that motivated Trump's executive order.
AI and Job Displacement: Data vs. Narrative
RAMP Study: AI Adoption Correlates with Hiring Growth
A study of over 21,000 US firms found that companies with highest AI spending grew fastest and increased headcount roughly 10% in two years post-adoption. Entry-level headcount rose even faster at 12%. Non-adopters saw flat headcount. No current data supports job loss narrative.
High AI adopters
10 % headcount growth
Entry-level roles (high adopters)
12 % headcount growth
Non-adopters
0 % headcount growth (flat)
Headcount growth by AI adoption intensity (2-year post-adoption)
Job Displacement vs. Job Loss Distinction
Certain jobs will be displaced (customer support, data entry, cab driving), but overall employment grows because AI creates new roles and productivity gains. The media conflates displacement with net job loss, a false equivalence. Displaced workers may transition to higher-value roles.
Specific Displacement Categories
Jobs most vulnerable to near-term displacement: customer support (AI handles tier-1 inquiries), data entry and business process outsourcing, cab driving (Waymo at 3,000 cars, Tesla FSD expanding), package sorting (Figure AI and Optimus robots). These are low-skill, repetitive tasks.
1
Customer support (tier 1)
High risk
2
Data entry and BPO
High risk
3
Cab driving
Medium-high risk
4
Package sorting
Medium-high risk
Job categories most vulnerable to AI displacement
Human-in-the-Loop Premium
As automation increases, human interaction becomes a premium service. Companies like Klarna initially tried full AI customer service but reverted to human agents for brand trust. Bartenders, massage therapists, and personal drivers command higher wages because humans value human connection over pure efficiency.
California's Fiscal Crisis and Hidden Liabilities
California Budget Ballooning: 65% Growth in 6 Years
California's state budget grew from $215 billion in 2019 to $355 billion in 2024—a 65% increase. Governor Newsom claims a balanced budget, but this masks $20-40 billion in debt that the state pencils away through accounting tricks, creating hidden liabilities.
2019 Budget
$215 billion
2024 Budget
$355 billion
California state budget growth (65% in 6 years)
Revenue Concentration: Top 1% Pays Half of Income Tax
California's top 1% (150,000 people) pays $70 billion of the state's $142 billion personal income tax revenue—50% of all income tax. The top 1,000 people alone pay $22 billion annually (11% of state income). This extreme concentration creates massive vulnerability to wealth exodus.
Top 1% (150k people) 50%
Remaining 99% 50%
California personal income tax distribution
Corporate Tax Exodus: 15 Fortune 500 Companies Left Since 2019
At least 15 Fortune 500 companies have relocated headquarters out of California since 2019, along with 2,100 mid-to-large companies. California's corporate tax rate is 8.9%, compared to Texas (0%), Tennessee (6.5%), and Florida (5.5%). This exodus has cost the state 5% of jobs.
California
8.9 % corporate tax
Florida
5.5 % corporate tax
Tennessee
6.5 % corporate tax
Texas
0 % corporate tax
Corporate tax rates: California vs. competitor states
Personal Income Exodus: 1-1.5% Annual AGI Drain
California loses 1-1.5% of adjusted gross income (AGI) annually as high earners leave. Over 10 years, this compounds to 15% of the state's income base leaving. The proposed billionaire tax will accelerate this exodus further.
1-1.5%
Annual AGI exodus from California
High-earner income drain per year
New Taxes on Middle Class: Software Sales Tax and Healthcare Tax
California introduced an 8% sales tax on software (Gmail, ChatGPT, Microsoft Word subscriptions), projected to raise $1 billion annually. A new corporate healthcare insurance tax will raise $2 billion. These shift the tax burden from wealthy to average Californians as the state spirals.
Software sales tax
1 $ billion annual revenue
Healthcare insurance tax
2 $ billion annual revenue
New California taxes targeting middle class
Unfunded Liabilities: $1.5-2 Trillion Hidden Debt
California has $1.4 trillion in public debt (state $500B, local $800B). Unfunded pension liabilities are officially $664 billion but estimated at $1.5 trillion. Retiree healthcare obligations add $175 billion. Total hidden liabilities: $1.5-2 trillion. Pension obligations rank senior to bonds under the California rule.
Reported pension liabilities
664 $ billion
Estimated true pension liabilities
1500 $ billion
Retiree healthcare obligations
175 $ billion
California's hidden and unfunded liabilities
Projected Budget Deficits 2028-2029: $40 Billion Annually
Despite Newsom's claimed balanced budget, projections show $40 billion annual deficits in 2028-2029. The state is borrowing to cover current spending while liabilities accumulate, setting up a fiscal cliff.
$40 billion
Projected annual deficit 2028-2029
California's looming budget gap
Spending Inefficiency: California vs. Texas and Florida
Texas and Florida spend ~$5,000 per capita on services; California spends ~$9,000; New York City spends ~$13,000. Californians do not receive 2x the value for 2x the spending. Problems are worse, not better. This reflects incompetence and mismanagement.
Texas
5000 $ per capita
Florida
5000 $ per capita
California
9000 $ per capita
New York City
13000 $ per capita
Per capita government spending by state
Potential Federal Bailout and Union Crisis
If California defaults and requires federal bailout, red states will demand why they should bail out blue state mismanagement. This could trigger a constitutional crisis and calls for secession from the union. A federal wealth tax or socialism intensification would likely follow.
Worth quoting
"Why would anybody do this when there are alternatives?"
— Alex Karp, Palantir CEO, at [1:36]
"If you partner with any of these people, they will slit your throat and take your business wholesale."
— Jason Calacanis, at [20:00]
"You cannot rent intelligence from the same place that rents it to your competitor."
— David Sax, at [24:34]
Made with Glimpse by Wozart
glimpse.wozart.com/v/yqzpwvoa
Share this infographic
Read this infographic as text

AI Sovereignty, Anthropic's Betrayal, SCOTUS Birthright Ruling, California's Fiscal Collapse

Summary of the video “AI Sovereignty Wars, Palantir-Nvidia Deal, SCOTUS Birthright Ruling, Newsom’s CA Budget Lie” by All-In Podcast.

Palantir and Nvidia partner on sovereign AI to keep government data domestic, exposing how frontier labs like Anthropic exploit enterprise data to build competing products. SCOTUS upholds birthright citizenship 6-3. California's $351B budget masks $20-40B in hidden debt, massive tax exodus, and $1.5-2T in unfunded liabilities, threatening state solvency.

AI Sovereignty and the Frontier Lab Trap

Palantir-Nvidia Sovereign AI Partnership

Palantir and Nvidia announced a partnership to build a custom frontier-quality AI model for US government agencies, where the government owns the hardware, data, and model weights. This addresses concerns that proprietary data shared with frontier labs gets exploited to build competing products.

Anthropic's Vertical Integration Playbook

Anthropic has systematically launched competing products in categories where customers built successful applications on top of Claude: Claude Design (competing with Figma), Claude Code (competing with Cursor), Claude Security, Claude Legal, Claude Financial. This mirrors Microsoft's and Google's historical playbook of using platform dominance to capture lucrative verticals.

Figma's Cautionary Tale

Anthropic's chief product officer sat on Figma's board and did not resign until 3 days before Anthropic launched Claude Design, a direct competitor. Figma's stock fell 50% in the year following, illustrating the risk enterprises face when sharing data with frontier labs.

Enterprise Cost and Performance: Open Source vs. Claude

8090's testing showed that wrapping Claude with their software factory harness achieved 1.4x cheaper and 1.5x faster results than Claude alone. Using an open-source frontier model with their harness was 16.4x cheaper, though 3x slower—a trade-off many enterprises find acceptable for massive cost savings and data sovereignty.

Intelligence Sovereignty vs. Privacy

Intelligence sovereignty means enterprises retain control over which AI analyzes their data and how it interprets their world—distinct from privacy, which is about preventing data access. It ensures competitors cannot use shared data to build better products against you.

The Distributed Inference Future

Enterprises are shifting from a hub-and-spoke cloud model to a distributed model: large hubs for foundational training, medium hubs for enterprise fine-tuning on proprietary data, and distributed spokes for local inference. This allows companies to run workflows on-premises, avoiding cloud dependency and data leakage.

Nvidia's Open-Source Strategy Shift

Nvidia has historically downplayed its open-source Neotron model to avoid alarming major customers. After OpenAI, Anthropic, and AMD announced competing chips and Elon signaled his own fab, Nvidia took the gloves off and began promoting Neotron as a competitive alternative to closed models, positioning itself as a full-stack provider.

Why Chip and App Companies Want Model Competition

Chip makers (Nvidia), applications (Palantir), and enterprises all benefit from a competitive model layer because it prevents lock-in and monopsony pricing. Only Anthropic and OpenAI want duopoly dominance. A healthy competitive model layer ensures diverse buyers for chips and prevents any single model provider from controlling enterprise destiny.

Token Price Deflation Ahead

With open-source competition intensifying, token costs are projected to fall 90% annually for the next 3 years, making enterprise AI dramatically cheaper and enabling companies to run wasteful experiments locally without concern.

Anthropic Export Controls and Palace Intrigue

Mythos Jailbreak and Export Restrictions

Anthropic's Mythos model was found to have a jailbreak that bypassed safety guardrails when tested by Amazon. Dario Amodei had previously called Mythos a cyber weapon. The US Commerce Department issued export restrictions on Fable and Mythos on June 30, but lifted them on June 26 after Anthropic replaced Dario with co-founder Tom Brown as lead negotiator.

Three Conditions for Export Control Letter

The export control letter resulted from three specific conditions: (1) Dario publicly calling Mythos a cyber weapon, (2) Amazon reporting the jailbreak failure to Commerce, and (3) Dario refusing to roll back Fable until the jailbreak was fixed. Change any one condition and the letter would not have been issued.

Open-Source Models Are Not Chinese Once Forked

Once a model like DeepSeek or Qwen is open-sourced, it ceases to be inherently Chinese because enterprises can fork it, run it on US hardware in US data centers with no data leakage to China. Banning open-source imports would put the US at a disadvantage globally and subject American companies to a token tax.

SCOTUS Birthright Citizenship Ruling

Supreme Court Upholds Birthright Citizenship 6-3

Chief Justice Roberts, joined by the three liberal justices, ruled that the 14th Amendment guarantees automatic citizenship to children born in the US regardless of parental immigration status. Trump's executive order ending this was struck down. Justice Kavanaugh suggested Congress could still tighten birthright rules consistent with the Constitution.

Original Intent vs. Textual Reading Debate

The 14th Amendment was ratified in 1868 to ensure freed slaves and their descendants had citizenship rights after Dred Scott denied it. Intentionalists argue the framers never contemplated modern immigration; textualists read the plain language as granting citizenship to all born in the US subject to its jurisdiction.

Proposed Compromise: Residency-Based Citizenship

A middle-ground proposal would grant birthright citizenship only to children of legal residents (green card holders), not temporary visitors or illegal immigrants. This preserves citizenship for long-term residents while excluding birth tourism and casual visitors.

Birth Tourism Example: Dong Wan Lee Case

In 2019, Dong Wan Lee pleaded guilty to helping pregnant Chinese nationals enter the US under false pretenses so their babies would gain US citizenship. She served 10 months in prison. This case illustrates the birth tourism concern that motivated Trump's executive order.

AI and Job Displacement: Data vs. Narrative

RAMP Study: AI Adoption Correlates with Hiring Growth

A study of over 21,000 US firms found that companies with highest AI spending grew fastest and increased headcount roughly 10% in two years post-adoption. Entry-level headcount rose even faster at 12%. Non-adopters saw flat headcount. No current data supports job loss narrative.

Job Displacement vs. Job Loss Distinction

Certain jobs will be displaced (customer support, data entry, cab driving), but overall employment grows because AI creates new roles and productivity gains. The media conflates displacement with net job loss, a false equivalence. Displaced workers may transition to higher-value roles.

Specific Displacement Categories

Jobs most vulnerable to near-term displacement: customer support (AI handles tier-1 inquiries), data entry and business process outsourcing, cab driving (Waymo at 3,000 cars, Tesla FSD expanding), package sorting (Figure AI and Optimus robots). These are low-skill, repetitive tasks.

Human-in-the-Loop Premium

As automation increases, human interaction becomes a premium service. Companies like Klarna initially tried full AI customer service but reverted to human agents for brand trust. Bartenders, massage therapists, and personal drivers command higher wages because humans value human connection over pure efficiency.

California's Fiscal Crisis and Hidden Liabilities

California Budget Ballooning: 65% Growth in 6 Years

California's state budget grew from $215 billion in 2019 to $355 billion in 2024—a 65% increase. Governor Newsom claims a balanced budget, but this masks $20-40 billion in debt that the state pencils away through accounting tricks, creating hidden liabilities.

Revenue Concentration: Top 1% Pays Half of Income Tax

California's top 1% (150,000 people) pays $70 billion of the state's $142 billion personal income tax revenue—50% of all income tax. The top 1,000 people alone pay $22 billion annually (11% of state income). This extreme concentration creates massive vulnerability to wealth exodus.

Corporate Tax Exodus: 15 Fortune 500 Companies Left Since 2019

At least 15 Fortune 500 companies have relocated headquarters out of California since 2019, along with 2,100 mid-to-large companies. California's corporate tax rate is 8.9%, compared to Texas (0%), Tennessee (6.5%), and Florida (5.5%). This exodus has cost the state 5% of jobs.

Personal Income Exodus: 1-1.5% Annual AGI Drain

California loses 1-1.5% of adjusted gross income (AGI) annually as high earners leave. Over 10 years, this compounds to 15% of the state's income base leaving. The proposed billionaire tax will accelerate this exodus further.

New Taxes on Middle Class: Software Sales Tax and Healthcare Tax

California introduced an 8% sales tax on software (Gmail, ChatGPT, Microsoft Word subscriptions), projected to raise $1 billion annually. A new corporate healthcare insurance tax will raise $2 billion. These shift the tax burden from wealthy to average Californians as the state spirals.

Unfunded Liabilities: $1.5-2 Trillion Hidden Debt

California has $1.4 trillion in public debt (state $500B, local $800B). Unfunded pension liabilities are officially $664 billion but estimated at $1.5 trillion. Retiree healthcare obligations add $175 billion. Total hidden liabilities: $1.5-2 trillion. Pension obligations rank senior to bonds under the California rule.

Projected Budget Deficits 2028-2029: $40 Billion Annually

Despite Newsom's claimed balanced budget, projections show $40 billion annual deficits in 2028-2029. The state is borrowing to cover current spending while liabilities accumulate, setting up a fiscal cliff.

Spending Inefficiency: California vs. Texas and Florida

Texas and Florida spend ~$5,000 per capita on services; California spends ~$9,000; New York City spends ~$13,000. Californians do not receive 2x the value for 2x the spending. Problems are worse, not better. This reflects incompetence and mismanagement.

Potential Federal Bailout and Union Crisis

If California defaults and requires federal bailout, red states will demand why they should bail out blue state mismanagement. This could trigger a constitutional crisis and calls for secession from the union. A federal wealth tax or socialism intensification would likely follow.

Notable quotes

Why would anybody do this when there are alternatives? — Alex Karp, Palantir CEO
If you partner with any of these people, they will slit your throat and take your business wholesale. — Jason Calacanis
You cannot rent intelligence from the same place that rents it to your competitor. — David Sax

More like this