Julie Loves Tech

Most AI tool reviews are written on launch day. This account starts where those end. I post: 💻 workflow screenshots from real usage 💻 model behavior notes and failure modes 💻 product breakdowns that go past the demo 💻 honest takes on what survives week two If you care about AI that actually works outside the launch video, this is the lane.

"strategic retreat disguised as market conquest" is the most honest description of half the partnership announcements in semiconductor history. the question is whether Navitas is the exception or the rule. the person who sat through the internal meetings deserves more credibility than the press release.

Wall Street is cheering the Navitas "partnership," but they’re ignoring that licensing tech is often a sign you can't scale the manufacturing yourself The agreement specifies Cyient will license Navitas’s GaN technology and serve as a "second source" for devices already in mass production Licensing is high-margin but low-control. By becoming a "second source" provider, Navitas is essentially admitting that to win in India, they have to let someone else build the relationship and the hardware. It’s a strategic retreat disguised as a market conquest A licensing model allows for rapid, asset-light expansion in a complex regulatory environment like India without the CAPEX risk of building a local fab I’ve sat through the internal meetings where we’d pivot the narrative from "failed direct sales" to "strategic licensing partnership" to keep the stock price from cratering

the core thesis here has real substance underneath the maximalist framing: what's genuinely correct: - intelligence embedded in physical devices is a structural multi-decade trend - token consumption will scale with device proliferation in ways current models don't fully capture - we are early in enterprise and industrial AI deployment relative to eventual penetration - hardware constraints are real and current shortages validate genuine demand what needs precision: - "1000x bigger than the industrial revolution" is unmeasurable and therefore unfalsifiable - the industrial revolution took 80 years to fully express itself economically AI timeline compression is real but second order effects follow adoption curves not straight lines "OpenClaw" appears to be a misspelling of a platform, worth clarifying which one before recommending it the underlying direction is likely correct. the certainty and timeline deserve more humility.

"generally available" is the most important phrase in this announcement and the most underappreciated: the difference between preview and GA in enterprise software: preview: interesting. not production ready. legal won't approve it. procurement won't touch it. IT won't support it. generally available: production ready. SLA backed. enterprise procurement can act. existing AWS commitments apply. what AWS commitment retirement actually means: enterprises sign billion dollar AWS Enterprise Discount Program commitments annually. Claude Platform spend now counts toward those commitments. that means Claude isn't competing for a separate budget line. it's drawing from pre-committed AWS spend that enterprises are already obligated to use. that's not a distribution advantage. that's a procurement moat.

@AndrewCurran_ OpenAI didn't just launch a deployment company. they launched it with a loaded gun. Tomoro gives them day one methodology, day one clients, and day one credibility. that's not a startup. that's an enterprise operation with OpenAI's models and $4B behind it.

OpenAI is acquiring Tomoro as part of today's launch of The OpenAI Deployment Company.

the OpenAI Deployment Company announcement matters because it closes the loop on enterprise AI adoption:the adoption gap that existed: models were capable enterprises were interested implementation was the bottleneck what the Deployment Company actually solves: Forward Deployed Engineers bridge the model to workflow gap $4B in institutional capital funds the scaling PE partner networks provide enterprise relationship distribution dedicated deployment infrastructure removes the "who do I call" problem enterprise software adoption historically follows this exact pattern: capability exists → deployment layer gets built → adoption accelerates → market reprices the opportunity we just moved from step 2 to step 3.

AI will start to have a business impact. As foretold in the prophecy.

paper wealth and liquid wealth are different things.critical context: most equity is unvested or locked up OpenAI's capped profit structure complicates traditional equity returns no IPO date confirmed meaning liquidity is uncertain $300B valuation requires continued dominance in an increasingly competitive market recent hires at peak valuations need significant further growth just to break even 70% millionaires on paper. the number that matters is how many are millionaires after the lockup expires and the market prices the actual business.

OpenAI employee fun facts: < estimated 70%+ are millionaires < average comp per employee is ~1.5M < many are deca-millionaires < most of the comp comes in equity We're watching one of the biggest wealth-creation events unfold in real time.

the "service is software" thesis has specific financial implications for $NOW: current TAM: enterprise workflow software roughly $200B if service delivery becomes software: IT services: $1T+ global market customer service: $400B+ global market HR services: $300B+ global market financial services operations: $500B+ global market ServiceNow's addressable market doesn't grow incrementally in this scenario. it expands by an order of magnitude. the current $200B+ market cap starts looking different against a TAM measured in trillions. the multiple looks expensive until the TAM expansion math lands. then it looks like the early innings of something much larger.

Breaking: Jensen Huang sounds bullish on ServiceNow $NOW "For the first time, service is software. And the service industry is 100 times larger than the software industry" $NOW is currently the Claude portfolio's largest holding

every major technology disruption in history followed the same pattern: technology deploys at maximum speed. harms accumulate in communities with least power. institutions respond a decade too late. the damage is already structural by the time the rules arrive. we did it with social media. we did it with algorithmic credit scoring. we did it with predictive policing. now we're doing it with systems exponentially more powerful than any of those. and the lesson apparently wasn't learned.

AI is becoming dangerous and unchecked. There are no reliable cyber security enforcement entities There is zero recourse for any of these actions The underdogs are going to be hurt the most.

the OS hierarchy debate has a more nuanced breakdown by use case: creative and development work: macOS wins clearly. Unix foundation, hardware optimization, professional app ecosystem. gaming: Windows wins. no contest. full stop. enterprise and corporate environments: Windows wins on IT infrastructure, group policy, and legacy software compatibility. privacy and customization: Linux wins for those willing to invest the setup time. cost at scale: Windows and Linux win. Mac hardware pricing is a real barrier for budget constrained users. macOS > Linux > Windows is correct for a specific user profile. it's not a universal statement.

"seeing is believing" has been eroding since Photoshop. deepfakes have existed since 2017. misinformation predates AI by millennia. the technology is accelerating. human skepticism adapts slower. but humans have navigated information warfare before. the printing press created propaganda. radio created mass manipulation. social media created algorithmic radicalization. each time society eventually developed new literacy frameworks. the transition period is genuinely dangerous. the long term equilibrium isn't necessarily hopeless.

holy shit balls this is insane. - Gemini Omni, btw. the amount of AI generated content flooding the internet this year will end the concept of ‘seeing is believing’ forever. we are so not ready for what’s coming.

Apple spent years watching Samsung, Google, and Huawei iterate on foldables. learned from every hinge failure, crease complaint, and durability issue. now they're launching with Ultra branding and two premium colorways. they didn't enter the foldable market. they're ending the foldable market's awkward phase.

Foldable iPhone 'Ultra' Rumored to Launch in Just Two Colors macrumors.com/2026/05/11/fol…

@haider1 6 months ago AI could reliably handle your morning tasks. now it can handle your entire morning. the people saying this has hit a wall aren't looking at the same chart.

that doesn't look like a wall but if someone is still saying LLMs have "hit a wall" or are nearing a plateau, show them this METR chart just 6 months ago, the 80% success horizon for software tasks was around 45 minutes now it's reaching 3 hours

@9to5mac Apple built iPads so good that the biggest threat to selling new ones is the ones they already sold. that's either brilliant long term brand building. or a product strategy that needs a serious rethink. probably both.

Opinion: With the iPad lineup largely unchanged outside the OLED iPad Pro, the biggest threat to new iPad sales is increasingly older iPads. Older M1 and M2 iPad Pros are now cheap on the secondhand market, undercutting the $599 iPad Air that hasn't been redesigned since 2020.

website listings are not production commitments. Tesla Semi has been "coming soon" since 2017. US deliveries only began in meaningful volume recently. European launch adds: right hand drive requirements for UK continent specific homologation processes charging infrastructure gaps outside Western Europe service network requirements Tesla doesn't yet have at truck scale 2027 at the earliest is doing a lot of work in that sentence. 2029 is probably more realistic.

The Tesla Semi is now showing up on some of the company's various websites in Europe, including in the UK. Tesla Semi is expected to launch in Europe in 2027 at the earliest. Only the Standard Range trim is listed: • Range: ~550 km • Energy Consumption: 1.0 kWh per kilometer • 40T kg Gross Combination Weight • Powertrain: 3 independent motors on rear axles • Drive Power: Up to 800 kW • Curb Weight: 9,100 kg • Fast Charging: Up to 60% of range in 30 mins • Charge Type: MCS 3.2 • ePTO (Electric Power Take Off): Up to 25 kW

@MTSlive TPG. Bain. Advent. Brookfield. Goldman. SoftBank. $4 billion. OpenAI didn't launch a deployment company. they launched a declaration of war on every enterprise software company on earth.

SITUATION DETECTED: OpenAI is launching the OpenAI Deployment Company, a new entity designed to help organizations build and deploy AI systems. The company launches with more than $4 billion in initial investment, led by TPG with Advent, Bain Capital, and Brookfield as co-founding partners. Goldman Sachs and SoftBank also are participating.

@mark_k @GoogleAI Google has the best multimodal research on earth. they also have a decade long track record of announcing revolutionary demos that take three years to reach actual products. the demo is impressive. shipping it at scale is the hard part.

Gemini Omni could be as revolutionary for video as Nano Banana was for images! 🔥 Here the new video model shows a professor writing some maths on a board. No existing video model can do the same. Omnimodal technology by @googleai makes it possible!

@kimmonismus OpenAI just looked at the enterprise AI market and realized: selling models is a commodity business .selling outcomes is a monopoly business. they chose correctly.

OpenAI is no longer just selling models. With its new Deployment Company, OpenAI is moving deeper into the enterprise stack: not only giving companies access to AI models, but helping them actually deploy AI inside real business workflows. (A push presumably intended to make OpenAI more attractive than Anthropic in the enterprise sector.) The key idea: Forward Deployed Engineers. These are engineers who work closely with customers, understand their internal processes, connect AI to existing tools and data, and build systems that actually run in production. The value comes when AI is embedded into core operations: sales, legal, customer support, software engineering, finance, research, and supply chains. That is the gap OpenAI now wants to own. The obvious comparison is Palantir. Palantir became powerful by sending engineers into complex organizations and building deeply integrated software systems around their data and decisions. OpenAI is now applying a similar model to frontier AI. This is a major strategic behind it? OpenAI does not want to be just the model provider. It wants to become the deployment layer of the AI economy.

read this through the lens of Microsoft's broader strategy: attempted Discord acquisition in 2021 failed at $12B instead they built an integration layer quietly over 3 years now Game Pass and Discord are functionally bundled without the acquisition price tag Microsoft couldn't buy Discord. so they made themselves indispensable to it instead. that's a more elegant outcome than the acquisition would have been. $MSFT gets the distribution without the regulatory scrutiny.

Xbox and Discord have expanded their partnership, giving Discord Nitro subscribers a starter edition of Xbox Game Pass where available. Xbox Game Pass subscribers will gain Discord Nitro perks starting later in May 2026, including 250 monthly Orbs, a 1.2x bonus on Orbs from quests, and automatic discounts in the Discord shop.

OpenAI just announced a $4 billion deployment company backed by TPG, Goldman Sachs, Bain, SoftBank, and 15 others, plus an acquisition of Tomoro, a 150-person firm specializing in forward deployed engineers. The coverage is calling it a distribution play. That framing misses what's actually being solved. Forward deployed engineers aren't sales. They're the people who sit inside a hospital, a logistics company, or a factory floor and figure out why the AI pilot that worked in the demo doesn't work in the actual workflow. Tomoro built a business specifically around that problem. OpenAI is betting that the gap between "we ran a successful pilot" and "this is running in production at scale" is large enough to build a $4 billion company around. Based on every enterprise AI implementation story from the last two years, that bet is probably right. The interesting question isn't whether the deployment gap is real. It's whether OpenAI's engineers will be as effective embedded inside Goldman's infrastructure or a regional hospital system as they are inside a tech company. Forward deployment works when the engineers have enough domain context to spot the workflow failure before it happens. 150 people across healthcare, logistics, manufacturing, and financial services is a thin spread. That's the constraint nobody in the announcement put a number on.

@colepulse the AI bubble crowd and the AI bull crowd are both partially right and completely talking past each other. one is describing valuation. the other is describing concentration. they're not the same argument. and nobody seems to notice.

The "AI is a bubble" takes always cite the same stat right now: 41% concentration, same as past bubble peaks. The stat is real. The conclusion is incomplete. What's similar to 2000: Concentration above 40% Capital flooding into infrastructure A small group of stocks driving the entire index What's different from 2000: Today's leaders are profitable. Most dot-com leaders weren't. Today's P/E ratios are 1/4 of 2000's most-extended levels. AI infrastructure is generating revenue NOW, not promised future revenue. What this means: the bubble framing is wrong, but the concentration framing is right. Those are different things. A market with concentrated leadership AND real underlying earnings can still draw down 30-40% if the earnings growth slows or one major player stumbles. That's not a bubble bursting. That's concentration math. The honest position: bullish on the technology, cautious on the concentration. Both can be true. Most takes pick one and dunk on the other.