LT Weekly: April 17, 2026

It was another tough quarter for Netflix. While Q1 was essentially in-line with lowered expectations when excluding the WBD termination fee windfall, Q2 guidance fell below Street projections. With that said, mgmt still believes that it is on track to hit the FY2026 outlook, albeit back-end loaded with operating margins.

On the plus side, the recent price hike has been well received, and their price / value ratio remains compelling. In the US, Netflix subs are paying “the least per hour of viewing compared to other SVOD offerings” and “in some case, you’d have to pay two times per hour to get a competitive service.”

Investors have also been very focused on the Co’s engagement trends and, while not transparent, mgmt’s KPI of “member quality” hit an all-time high. More traditional measures like view hours were also up and the Co saw positive retention across regions y/y. All of this supports the argument that incremental products like gaming (still viewed as a huge opportunity), podcasts, and live events are enhancing stickiness rather than cannibalizing core viewing.

Speaking about live content, it was interesting to hear Netflix confirm that it is currently in discussions with the NFL, but still within a sports strategy of tent-pole events which has not changed.

The advertising business is on track to double y/y in 2026, and programmatic is expected to grow to over 50% of non-live advertising revenue over time.

AI also remains a cornerstone to the narrative and NFLX thinks they will be leading the charge with AI production, and their recent acquisition of InterPositive will bolster their capabilities on this front.

Lastly, Reed Hastings’ decision to not stand for Board re-election post his current term (expires at the Annual Meeting in June) also stood out. It is the end of an era.

Overall, investors were hoping for more, but little has changed with the yearly outlook and in our mind it is all about mgmt executing on the playbook and new initiatives looking ahead.

-> NFLX shares fell -10% in reaction but the stock is still up +4% YTD

Q1 Was Essentially ~In-Line Though WBD Termination Fee Was A Windfall To Adj EPS & FCF

• Rev BEAT cons by +0.6% and grew by +16% y/y (+14% y/y FXN)
  • Growth was driven primarily by membership growth, higher pricing, and increased ad revenue, with additional upside from favorable F/X and stronger-than-expected subs

• Op margin MISSED cons (32.3% vs cons 32.4%)
  • +60bps y/y due to op leverage

• Adj EPS BEAT cons by +61.8%
  • Driven by higher op income & a $2.88bn termination fee related to WBD transaction

• FCF BEAT cons by +67%
  • Driven by stronger operating performance and the termination fee

• Repurchases resumed following the termination of the Warner Bros. transaction…Q1 buyback was $1.3bn, leaving $6.8bn remaining on the authorization

Q2 Guidance Disappointed Investors BUT Mgmt Reiterated The Full Year 2026 Outlook…Margin Growth Remains H2 Weighted

• Q2 guidance MISSED cons… still implies moderating growth and margin pressure
  • Rev guidance of $12.57bn MISSED cons by -0.1%…implies +13.5% y/y growth (+12% FXN) vs cons of +14.1%
    • Growth reflects cont’d membership and pricing growth but decelerating vs Q1
  • Op margin guidance of 32.6% MISSED cons 33.4% but is up from 34.1% y/y and reflects higher content amortization
  • EPS guidance of $0.78 MISSED cons of $0.84
    • Reflects normalization vs Q1’s non-recurring benefit

• 2026 guidance UNCHANGED
  • Rev guidance at the mid-pt is +0.3% ahead of cons and implies +12-14% y/y growth (+11-13% FXN), vs +16% y/y in 2025
    • Growth will be driven by membership growth, pricing, and a projected ~2x increase in ad revenue
  • Op margin guidance of 31.5% (includes ~$275mn of acq-related expenses) remains below cons 32.7%, but is up from 29.5% in 2025
    • Margin progression remains back-end loaded due to higher content amortization growth in H1, with moderation in H2

• 2026 FCF guidance RAISED to ~$12.5bn (vs prior $11bn), driven by the WBD termination fee and strong underlying cash generation

“Member Quality” KPI Hits An All-Time Highs Based On Mgmt’s Data

• Total Q1 view hours growth was consistent with H2’25 trends despite external competition (e.g., Olympics) but mgmt. reiterates how that is not the only engagement metric
• Their primary engagement KPI of “member quality” hit another all-time high…retention improved across all regions y/y
  • They don’t disclose their “Member Quality” calculation for competitive reasons

Growth In The Advertising Business Remains ON TRACK…Programmatic Is Also Expected To Increase To Over 50% Of Non-Live Ads Mix Over Time

• Mgmt reiterated expectation to ~2x ad revenue y/y to ~$3bn in 2026
• The Co’s advertiser ecosystem is scaling rapidly
  • The advertiser base grew +70% y/y to 4,000+ and expect further expansion
  • But NFLX is still concentrated in the top advertiser accounts

• Their strategic path for ad revenue growth…
  • Have been making it easier for advertisers to buy on their platform
  • Added more DSP integrations which also provides more ways to buy
  • Improving go-to-mkt capabilities
  • Increasing the sales force
  • Continuing to build ad products
  • Leverage AI w/in the Netflix ad suite which will make it easier to design new creative formats, custom ads, improved contextual relevance, etc

• Programmatic ad buying will increase…it is expected to exceed 50% of non-live ads revenue over time

• Ad tier traction remains strong
  • >60% of sign-ups in ads markets are coming from ad-supported plan

The Co Maintained Its Cash Content Spend Ratio

• Continue to expect an annual cash content spend-to-amortization ratio of ~1.1x
• The vast majority of their content budget remains on core series and films…highlights in the slate:
  • Films
    • Here Comes the Flood with Denzel Washington
    • Greta Gerwig’s Narnia
  • Series
    • Will Ferrell’s The Hawk
    • One Hundred Years of Solitude S2
    • Lupin Part 4

Live Content Is Still Small BUT Demonstrating Disproportionate Impact…Discussions With The NFL Are Underway

• The Co aired 70+ live events in Q1, compared to 200+ across all of FY2025, ie, already accounting for ~35% of last year’s total…Q1 successes include:
  • World Baseball Classic which was exclusively for members in Japan was a huge hit
    • Delivered 31.4mn viewers
    • 3rd most-watched program ever on Netflix in Japan
    • Led to the largest day of sign-ups in the country (Japan was the largest contributor to member growth in Q1)
  • BTS The Comeback Live also had a strong showing
    • Delivered 18.4mn global viewers
    • Reached the weekly Top 10 in 80 countries
  • Coming up…UK fight between Tyson Fury and Anthony Joshua

• The Cos sports strategy is “unchanged” as they are most interested in those “big breakthrough events, less so in the regular season packages”
  • Ramping up sports events globally and locally
    • Multiyear deal with CONCACAF for rights in Mexico
    • The Women’s World Cup in US and Canada
    • The M&A event with Ronda Rousey and Carano

• NFLX is “in discussions” with the NFL right now as they “think there’s an opportunity to expand the relationship”
  • “We’ve learned a lot about how what works and how to value the NFL and live generally over the last couple of years”
  • “And this is going to inform how we have those discussions and help us be much even more disciplined”

Bullish On The Expansion Into Gaming & Podcasts Are Showing Incremental Engagement

• Continue to see gaming as a “significant” market opportunity…”we’re still really just scratching the surface today in terms of what we can ultimately do in this space”
  • Launched a new “Playground” kids app
    • ~10% of kids’ profiles have played NFLX games and almost 50% of kids’ profiles view content on mobile and table
    • Expect to add more kids’ games in “the coming months”
  • Seeing “promising traction” for their cloud TV games
  • “We believe game play can have a positive impact on member retention”

• Podcasts are showing early signs of incremental engagement (not substitution)…this can be seen given the Co is over-indexing on daytime viewing (which is normally lighter) and higher mobile usage

“Netflix Is Going To Remain At The Forefront In The Exploration And Innovation Of AI In The Creative Process”

• The Co focuses generative AI investment where it has “differential or unique capability to deliver returns to the business”
  • Content: “Content production is a big one” and “member experience is another big one”
  • User experience: “We’ve been in personalization and recommendation for two decades” but still see “tremendous room and opportunity to make it even better by leveraging some of these newer technologies”
    • New model architectures in recommendation systems are already “driving increased engagement with the service” and act as a “force multiplier” on content investments

• On the content side, “it takes a great artist to make great art and AI won’t change that”
  • AI will give those artists better tools to bring those visions to life
  • The tools will help with “references, previsualization, visual effects, sequence prep, shot planning”

• The acq of InterPositive accelerates NFLX’s Gen AI capabilities b/c it’s a proprietary tech that was created specifically for filmmakers
  • Even though ownership of InerPositive is new, there have been a “bunch of interest” with their creators

Several Other Key Points

• Netflix still has a big TAM for future growth …they currently have:
  • <45% penetration of broadband households
  • ~5% global TV view share globally
  • ~7% share of their addressable revenue pool

• Price increases have “gone well” and just annc’d price adjustments in Spain
  • Subscribers are still getting great value:
    • In the US, Netflix subs are paying “the least per hour of viewing compared to other SVOD offerings”
    • “In some case, you’d have to pay two times per hour to get a competitive service”

• The Co is launching vertical video at the end of the month

Just two weeks after activist investor Irenic Capital Mgmt’s public letter to Snap CEO Evan Spiegel advocating for employee cuts, changes to stock comp frameworks, and increased use of AI for internal efficiencies (plus several other proposed actions that it thought would get the stock to $26.37/share), Spiegel officially announced employee and cost guidance changes this week. 16% of the workforce is being eliminated, which will lead to $500mn+ in annualized savings. The Co also reduced its 2026 cost guidance for adj operating expenses and stock-based comp. Preliminary Q1 adj EBITDA is also 26% ahead of current consensus, with Q1 revenue trends on track (was a sigh of relief post last qtr’s trends).

Given Irenic’s argument that Snap’s stock based comp (SBC) per employee is too high, we did a quick calculation using Snap’s new lower headcount and the new SBC 2026 guidance, and SBC/employee actually goes from $193k in 2025 to an estimated $237k in 2026, so some investors could argue that the SBC cuts could have been higher.

Regardless, the Co’s cost base is coming down as Snap navigates being “squeezed between giants with enormous resources and nimble startups moving fast.”  It is focused on its structural pivot toward “profitable growth” which mgmt. began talking about on their Q3:25 earnings call in November.

We expect additional color on this announcement during the Co’s May 6^th earnings call, along with updates on the future direction of the Spectacles business (Irenic was pushing the Company to either sell it or shut it down), the Perplexity deal (which is reportedly “dead” – link), and any competitive pressure from OpenAI’s new advertising launch.

See below for more of the key points in our mind from Snap’s update.

-> Snap shares rallied +6.5% on the back of the update and are up +53% from its $3.93 low on March 27^th, 2026; With that said, it is still down -39% from its 52-wk high of $10.40 back in July 2025

• Q1 preliminary numbers are ABOVE cons, especially on profitability:
  • Rev of $1.53bn was a slight +1% above cons (and is at the high end of the prior guidance of $1.50-1.53bn)
  • Adj EBITDA of $233mn BEAT cons by +26% (and prior guidance of $170-190mn)

• Cutting ~1,000 employees or ~16% of the full-time base (note that Irenic had advocated for ~1.5k total cuts…1k core + 500 Spectacles)
  • Est’d annualized cost savings is $500mn+ (based on H2:26)
  • Restructuring costs: ~$95-$130mn (the majority expected to be incurred in Q2)

• Mgmt updated the FY 2026 cost structure guidance, with a bias towards reductions:
  • Maintained FY guidance for:
    • Infrastructure costs: $1.6-1.65bn
    • All other adj cost of rev (ex-Infrastructure costs): 16–17% of rev
  • Lowered FY guidance for:
    • Adj operating expenses: Cut by $250mn to ~$2.75bn
    • Stock-based compensation: Cut by $150mn to ~$1.05bn

-> Back of the envelope, based on a new est’d employee count of 4,419 after the 16% reduction, and the new $1.05bn stock-based comp, that implies $237k SBC/employee which would be an increase from $193k in 2025; Aside from cuts to SBC/employee, Irenic was also pushing Snap to move from a purely time-based approach to 2/3 performance vested w/ $10 and $15 share price targets

Irenic’s previous analysis circulated on March 31^st

Source: Irenic Presentation

• Mgmt reiterated 60%+ gross margins guidance for 2026, which is slightly above cons 59% and up from 55% in 2025…how to get there:
  • Monetizable growth focus (higher-value geos / users)
  • Mix shift to higher-margin ad placements
  • Scaling subscriptions (Snapchat+ + direct rev streams)

• The Co has been “significantly scaling” AI across product + engineering workflows enabling teams to focus more on strategic priorities and move faster
  • 65%+ of new code is generated by AI
  • AI agents are delivering measurable productivity
    • 1mn+ support queries/month have been handled
    • 7,500+ bugs have been identified via code review agent

• Mgmt reiterated their long term strategic goals
  • 1bn MAUs
  • Revenue growth pillars
    • Accelerate direct response advertising share gains
    • Increase content engagement to drive monetization
    • Scale subscriptions + diversify revenue
  • Financial goals
    • Sustained EBITDA profitability + positive FCF
    • A clear path to net income profitability
  • Strategic positioning
    • Lead the next computing platform transition

Stanford University’s 2026 AI Index Report was one of the most comprehensive data reports that we have seen, maybe ever! The 423-page report (!) has AI data related to technical performance, infrastructure, environmental impact, jobs and workforce, investment levels, impact on science/medicine, and more. The authors pride themselves in being a 100% neutral party trying to provide the public, corporations, researchers, and governments data on which to come to their own strategies and conclusions about AI. Big picture though, the data shows that AI is scaling faster than systems around it can adapt.

It was hard to pick our Top 15 insights, given there are many, many, more, but see below. (One of most fascinating was that was that top models read analog clocks correctly just 50.1% of the time vs 90.1% for humans!).

The full report can be accessed HERE

• Top model performance is converging, w/ 4 companies now clustered within 25 Elo points (inspired by chess ratings) when rated against one another by human voting in the Arena Leaderboard and benchmark

• The AI gap between the US and China has closed…
  • In February 2025, DeepSeek-R1 briefly matched the top U.S. model, and as of March 2026 Anthropic’s top model leads by just 2.7%
  • China leads in publication volume, citations, and patent grants, while the U.S. retains higher-impact patents and produced 50 notable models in 2025 vs China’s 30
  • South Korea leads in AI patents per capita, and China’s share of the top 100 most-cited AI papers grew from 33 in 2021 to 41 in 2024

• The US has 10x as many datacenters (5,427) and consume more energy than any other country
  • Also, TSMC fabricates almost every leading AI chip, making the global AI hardware supply chain dependent on one foundry in Taiwan (though a TSMC-U.S. expansion began operations in 2025)

• Regarding AI’s environmental footprint in 2025:
  • Grok 4’s est’d training emissions reached 72,816 tons of CO₂ equivalent
  • AI data center power capacity rose to 29.6 GW which is comparable to NY state at peak demand
  • Annual GPT-4o inference water use alone may exceed the drinking water needs of 12mn people
  • See below for est’d data center electricity consumption by region through 2030

• AI models can win a gold medal at the International Mathematical Olympiad BUT CANNOT reliably tell time
  • Researchers call this jagged frontier of AI
  • Gemini Deep Think earned a gold medal at IMO
  • BUT top models read analog clocks correctly just 50.1% of the time vs 90.1% for humans (who failed this?!)

• Global corporate AI investment more than doubled y/y in 2025
  • Private investment grew fastest at 127.5% y/y and now accounts for 60% of the total
  • Generative AI investment grew more than 200% y/y and captured nearly half of all private AI funding

• The U.S. also led in entrepreneurial activity with 1,953 newly funded AI companies in 2025, more than 10x the next closest country
  • BUT the number of AI researchers and developers moving to the U.S. has dropped 89% since 2017, with an 80% decline in the last year alone
    • With that said, the U.S. is still home to more AI talent than any other country

• OpenAI’s compute spend towers over all other and incr’d significantly from 2024 to 2025, as did Anthropic’s

• Video generation models are starting to capture how objects behave
  • Google DeepMind’s Veo 3, tested across more than 18k generated videos, demonstrated abilities like simulating buoyancy and solving mazes without being trained on those task
  • See below image showing how visual generation has improved over time…

• AI companies grew less transparent this year
  • After rising on the Foundation Model Transparency Index from 37 to 58 between 2023 and 2024, the avg score dropped to 40 in 2025
  • Training code, parameter counts, dataset sizes, and training duration are no longer disclosed for several of the most resource-intensive systems, including those from OpenAI, Anthropic, and Google

• AI models struggle to tell the difference between knowledge and belief
  • When a false statement is presented as something another person believes, models handle it well
  • When the same false statement is presented as something a user believes, performance collapses

• AI models perform better in business domains like tax, mortgage processing, corporate finance, and legal reasoning BUT have challenges in fields with where high competency and reliability are required
  • Studies report gains of 14% to 15% in customer support, 26% in software development, and 50% in marketing output
  • In software development, where AI’s measured productivity gains are clearest, U.S. developers ages 22 to 25 saw employment fall nearly 20% from 2024, even as the headcount for older developers continues to grow

• Nearly two-thirds of Americans (64%) expect AI to lead to fewer jobs over the next 20 years, while only 5% expect more
  • Experts were less pessimistic (39% fewer, 19% more) but forecast far faster adoption

• Frontier models outperform human chemists on avg but CANNOT reproduce published research
  • On ChemBench, the best models surpass human expert averages across 2,700+ chemistry questions while struggling with basic tasks
  • On ReplicationBench, frontier models score below 20% on paper-scale replication in astrophysics
  • On UnivEarth, LLM agents answer Earth observation questions with 33% accuracy, and their code fails 58% of the time
  • On end-to-end scientific research tasks, the best AI agents score roughly half of what PhD experts achiev
  • On PaperArena, the best agent reaches 38.8% accuracy versus a PhD expert baseline of 83.5%. On BixBench, frontier models achieve roughly 17% accuracy on real-world bioinformatics analysis

• The US reported the lowest trust in its own govt to regulate AI responsibly of any country surveyed, at 31%
  • The global avg was 54%
  • Southeast Asian countries are leading (Singapore 81%, Indonesia 76%)
  • The EU is trusted more than the US or China to regulate AI effectively

One theme that has been gaining momentum is US state-level push back regarding data center expansion and there were a couple developments on that front this week.  In Maine, lawmakers on Tuesday approved a first-in-the-nation, statewide moratorium which would pause the approval of new data centers above 20 megawatts for 18 months, alongside a separate bill removing certain tax exemptions and a new study group to assess environmental and energy impacts. The Governor has 10 days to veto the legislation, sign it into law, or allow it to become law without her signature.

In Virginia, a recent Washington Post–Scar School poll shows support for new data centers falling from 69% in 2023 to 35% in 2026. This comes as Prince William County dropped its legal appeals tied to the proposed “Digital Gateway,” effectively ending years of efforts to revive a 2,100-acre hyperscale campus. The broader region already hosts one of the largest data center hubs in the world, with northern Virginia’s load exceeding 4,900 MW, while the Digital Gateway alone was planned at multi-gigawatt scale.

These updates are not isolated events. There is a long list of other states that have started their own processes dating back to the beginning of the year. More color on the Maine and Virginia updates are below, in addition to other state actions regarding data center expansions on a state-by-state basis …

• Maine could be the 1st state to pause new data center builds as lawmakers passed the first state temporary ban on Tuesday (link/link)
  • What would this legislation do? It would pause approvals of new data centers for 1.5 years that require more than 20 megawatts of power for 18 months; It would also establish a study group to examine the impact of such facilities and recommend legislative guardrails
    • Separate legislation, also approved, would make data centers ineligible for certain business tax exemptions
  • What is next? Governor Janet Mills has 10 days to veto the legislation, sign it into law, or allow it to become law without her signature (see more of amendment LD 307 HERE)
  • Why the push back against datacenters? Lawmakers are concerned about the facilities’ use of water and electricity, the risks of pollution and the potential to drive up household energy costs

• Also this week, in Virgina, support for new data centers fell from 69% in 2023 to 35% according to a Washington Post-Scar School poll (link/link)
  • This comes one day after Prince William County’s Board of Supervisors voted to drop its appeals defending the “Prince William Digital Gateway”
    • The decision ended years of legal effort to revive a 2,100-acre, 37-building campus that, at full buildout, would have been one of the largest data center developments in the country
  • More broadly, the region already anchors a large amount of hyperscaler capacity, as northern Virginia’s data center load exceeds 4,900 MW
    • The Digital Gateway alone was sized up for multi-gigawatt demand, across some 22mn sqft of building space and 14 dedicated electrical substations
  • Data Center Watch, a tracker run by 10a Labs, recorded 48 data center projects blocked or delayed across the US in 2025
    • This prevented ~$156bn in planned development, with 57 active opposition groups now operating in Virginia (more than any other state)

• Looking deeper, see below is a quick overview of the other states pursuing moratorium bills as of March 2026 (link)
  • Additional moratorium bills are still being discussed in Minnesota and Pennsylvania as well

AI players continue to roll out new models and tools at a rapid pace, while also becoming more explicit about how they see the competitive landscape and their positioning within it. This week, a leaked memo from OpenAI pulled back the curtain on how it’s thinking about its position within the AI landscape, emphasizing a shift toward a more integrated platform, taking direct shots at Anthropic, and sharing more nuanced views on its relationships with Microsoft and Amazon. Additionally, OpenAI followed Anthropic’s cybersecurity push last week and announced its own defensive-focused model, GPT-5.4-Cyber this week. Over at Anthropic, the Co rolled out an upgraded flagship model, Opus 4.7, and a new AI design tool, Claude Design, which lets users create visuals like prototypes, slides, one-pagers, and more using Claude (weighing on share of companies like Figma).

Finally, Microsoft, Meta, and Google each highlighted new efforts across agents, digital avatars, and robotics that we thought were notable.

See below for more of our takes.

OpenAI Shares Candid Thoughts On Anthropic, Microsoft, And Amazon in A Leaked Memo + Releases An AI Model For Defensive Cybersecurity

• A leaked internal memo from OpenAI’s Chief Revenue Officer Denise Dresser gave some interesting insight into the Co’s direction (full memo HERE/link): Key points below…
  • Confirms the existence of a new flagship model, codenamed “Spud”, which is described as “an important step in the intelligence foundation for the next generation of work,” focusing on complex reasoning and “intent recognition”
    • “Spud will make all of our key products significantly better. It expands the workflows we can own and gives customers another reason to consolidate around us”
    • This model is reportedly the engine behind OpenAI’s transition from a chat interface to a “Super App” capable of autonomous agency
  • Seemingly favoring AMZN over MSFT…Partnership w/ MSFT has been “foundational” BUT has limited their ability “to meet enterprises where they are – for many that’s [Amazon’s] Bedrock”
    • Since announcing a partnership w/ AMZN in Feb, inbound demand from customers for the offering has been “frankly staggering”
    • “We are firing on all cylinders to establish this as a scaled distribution channel”
  • Some strong words on Anthropic – “their story is built on fear, restriction, and the idea that a small group of elites should control AI”
    • How does OpenAI compare? “Our positive message will win over time: build powerful systems, put in the right safeguards, expand access, and help people do more”
  • Characterized Anthropic as a “single-product company in a platform war”
    • “Their coding focus gave them an early wedge…as AI spreads beyond developers into every team, workflow, and industry, that narrowness can become a real liability”
  • And OpenAI has a “real structural advantage” vs Anthropic
    • “Their strategic misstep to not acquire enough compute is showing up in the product. Customers feel it through throttling, weaker availability, and a less reliable experience. We saw the exponential compute curve earlier [and] acted on it faster”
  • They also allege that Anthropic is inflating its rev figures by ~$8bn: Anthropic is using gross accounting, recording the full price of Claude credits sold via AWS/Google as rev, while OpenAI uses net accounting, only recording what they actually keep after Microsoft takes its cut
    • That would put Anthropic’s true run rate at $22bn (vs the currently stated $30bn) and below OpenAI’s ~$25bn
  • Also outlined plans to “own deployment” w/ “DeployCo” to “turn product demand into repeatable enterprise transformation”
    • “The companies that win enterprise AI will not just have the best models. They will have the best ability to get those models deployed into real workflows, inside real organizations, with real measurable value. We should be the best in the world at that”
  • Overall… “we should stop thinking like a company with separate product lines. We should think like a platform company with multiple entry points and one integrated enterprise offering”

• OpenAI launched GPT‑4‑Cyber, an AI model for defensive cybersecurity + expanded its Trusted Access for Cyber (TAC) program (link/link)
  • GPT-5.4-Cyber is a variant of GPT-5.4 that is fine-tuned specifically for defensive cybersecurity work
    • It includes binary reverse engineering capabilities, which allow security professionals to analyze compiled software for malware, vulnerabilities, and security robustness without access to source code
    • B/c the model carries higher risk than standard deployments, the model will initially be rolled out on a limited basis to vetted security vendors, organizations and researchers
  • Scaling up OpenAI’s TAC program to thousands of verified individual defenders and hundreds of teams responsible for defending critical software
    • OpenAI is adding new tiers to its TAC program, which was launched in February, with higher levels of verification unlocking more powerful capabilities
    • Users approved for the highest tier will gain access to GPT-5.4-Cyber
    • The program’s expansion comes ahead of the release of more capable models in the coming months
  • As a reminder…this comes just a week after Anthropic annc’d Project Glasswing, a cybersecurity initiative built around Claude Mythos Preview

Anthropic Releases Claude Opus 4.7 + Launches Claude Design for AI-Powered Visual Creation

• Anthropic releases its new model Claude Opus 4.7 (link/link)
  • Key highlights –
    • “Notable improvement” over Opus 4.6 in software engineering: Users report being able to hand off their hardest coding work (the kind that previously needed close supervision) to Opus 4.7 with confidence
    • Also has “substantially better” vision: Can see images in greater resolution, and is more “tasteful” and “creative” when completing professional tasks, producing higher-quality interfaces, slides, and docs
    • “Substantially better” at following instructions: Where previous models interpreted instructions loosely or skipped parts entirely, Opus 4.7 takes the instructions literally
    • Is a more effective finance analyst than Opus 4.6, producing rigorous analyses and models, more professional presentations, and tighter integration across tasks
    • Better at using file system-based memory: It remembers important notes across long, multi-session work, and uses them to move on to new tasks that, as a result, need less up-front context
  • How does it compare to Anthropic’s most powerful model, Claude Mythos Preview? Its cyber capabilities are not as advanced as those of Mythos Preview; They are releasing Opus 4.7 with safeguards that automatically detect and block requests that indicate prohibited or high-risk cybersecurity uses
  • Availability: Available across all of Anthropic’s Claude products, its application programming interface and through cloud providers Microsoft, Google, and Amazon; The new model is the same price as Claude Opus 4.6

• On the back of that…Anthropic also intro’d Claude Design, a new Anthropic Labs product designed to simplify visual creation w/ AI (link/link)
  • What is it? An AI-powered design tools that enables users to collaborate with Claude to create polished visual work like prototypes, presentations, and marketing assets through simple conversational inputs
  • How does it work?
    • A user can describe what they need and Claude builds a first version
    • From there, they refine through conversation, inline comments, direct edits, or custom sliders (made by Claude) until it’s right
    • When given access, Claude can also apply the team’s design system to every project automatically, so the output is consistent with the rest of the Co’s designs
  • Collaboration is a major focus: Users can share designs within their organization, grant editing access, and work together in real time
  • Teams have been using Claude Design for…
    • Realistic prototypes
    • Product wireframes and mockups
    • Design explorations
    • Pitch decks and presentations
    • Marketing collateral
    • Frontier design
  • Claude Design is powered by Anthropic’s newly-released model, Claude Opus 4.7
  • Available in research preview for Claude Pro, Max, Team, and Enterprise subscribers

-> Shares of Figma and Adobe were down -6.8% and -1.5%, respectively, on the day of the release

A Couple Other Notable AI And AI-Adjacent Updates From Microsoft, Meta, and Google… 

• Microsoft is building OpenClaw-style agentic AI features within Copilot, aiming to make it more autonomous while addressing the security gaps in the open-source original (link/link/link)
  • “Exploring the potential of technologies like OpenClaw in an enterprise context”, including a team of always-on agents that would be able to work 24/7 on a users’ behalf within Microsoft 365 applications
    • The always-on version of Microsoft 365 Copilot could reportedly do things like monitor a user’s Outlook inbox and calendar and serve up a list of suggested tasks each day
    • Microsoft is also exploring OpenClaw-like agents tailored to certain roles, such as marketing, sales, and accounting, to “limit the permissions the agent needs,” siloing them from other parts of a biz
    • Another goal is to make 365 Copilot better at working within Microsoft apps in the background w/o needing direct oversight: For instance, reorganizing one tab of an Excel spreadsheet while users work on other tabs
  • Prioritizing enterprise customers, with enhanced access controls to reduce risks
    • Reportedly confident that it can implement “safer” versions of the tool
  • Timeline is still TBD: Engineers working on those features are reportedly optimistic that some will be ready in time for a preview at Microsoft’s Build developer conference in June, BUT the efforts are still nascent and could change

• Finally…Mark Zuckerberg is reportedly building an AI-powered 3D clone to interact with employees and replace him in meetings (link)
  • Zuckerberg is reportedly personally involved in training and testing his animated AI, which could offer conversation and feedback to employees
  • The character is reportedly being trained on Zuckerberg’s mannerisms, tone and publicly available statements, as well as his own recent thinking on Co strategies, so that employees might feel more connected to him through interactions with it

• Google introduces Gemini Robotics ER 1.6 that enhances reasoning to help robots navigate real-world tasks (link/link)
  • Is a “significant” upgrade to Google’s reasoning-first model that enables robots to understand their environments “with unprecedented precision”
    • Is capable of executing tasks by natively calling tools like Google Search to find information, vision-language-action models (VLAs) or any other 3P user-defined function
  • This model specializes in capabilities “critical” for robotics, including visual and spatial understanding, task planning and success detection Also unlocked a new capability – instrument reading: Enables robots to read complex gauges and sight glasses
  • Availability: Available now to developers via the Gemini API and Google AI Studio

In the ever-ongoing battle of social media platforms between TikTok, Instagram, and Snapchat, we thought that a recent Pew Research Center report offers a helpful step back into how teens are actually using and experiencing each platform. While entertainment is the common thread across all three, the roles they play are starting to look pretty different. TikTok has become the go-to for discovery, whether it is content, products, or even news, while Snapchat is much more about staying connected, with teens using it to keep up with friends and communicate throughout the day.

At the same time, most teens say their overall experience on social media is positive or at least neutral, even if there are some clear trade-offs. Concerns around time spent show up most on TikTok, along with more negative views on sleep and productivity. Snapchat, despite its strength in communication, stands out for higher rates of reported harassment. And across all of this, parents tend to see things more negatively than teens, especially when it comes to how much time is being spent and the broader impact of these platforms.

The report was chock full of interesting stats….see below for the ones we found most interesting (and if you would like to access the full report, you can do so HERE).

Platform Use Cases For Teens Are Distinct…TikTok Leads in Discovery, Snapchat in Connection, While Entertainment Dominates Overall

• Entertainment is the top use case across all three platforms, cited by roughly 9 in 10 or more users on each
  • TikTok dominates on entertainment…96% cite it as a reason, with 82% calling it a major reason
  • Instagram and Snapchat users also cite entertainment, but at slightly lower rates (91% and 87% respectively)

• Staying connected with friends and family is the most prevalent use case for Snapchat
  • Two-thirds of Snapchat users say keeping up with friends and family is a major reason they use the app, higher than Instagram (55%) or TikTok (33%)

• TikTok leads in product discovery and news consumption
  • ~6 in 10 TikTok users go there for product reviews or recommendations, vs about half on Instagram and far fewer on Snapchat
  • ~4 in 10 TikTok and Instagram users cite news as a reason they use the platform; Only ~1 in 4 Snapchat users say the same

• Very few teens use any of these platforms primarily for politics, though TikTok and Instagram edge out Snapchat on this metric

TikTok Draws The Most Time-Spent Concern Amongst Both Teens And Parents

• TikTok users are most likely to feel they spend too much time on the app
  • 28% of TikTok users say they spend too much time on it, vs 19% each for Instagram and Snapchat
  • Still, the most common response across all three platforms is “about right”… roughly 6 in 10 or more say this for each app

• A significant parent-teen perception gap exists, widest for TikTok
  • 44% of parents say their teen spends too much time on TikTok vs only 28% of teens who say the same (a 16-pt gap)
  • The same pattern holds for Snapchat (30% parents vs. 19% teens) and Instagram (27% vs. 19%)

• Higher-income parents are more likely to view their teen’s usage as excessive
  • Among parents of TikTok users, 47% of those in households earning $75k+ say their teens spend too much time on the app vs 36% of those earning < $75k
  • A similar income-based pattern holds for Snapchat but no significant income difference for Instagram

Platform Impacts Diverge…TikTok Skews Negative On Sleep/Productivity While Snapchat Leads On Friendships

• TikTok stands out most for negative effects on sleep and productivity
  • 37% of TikTok users say it hurts the amount of sleep they get vs ~24–26% for Instagram and Snapchat
  • 29% of TikTok users say it hurts their productivity, compared to ~21–22% on the other two platforms

• Snapchat is the standout platform for positive effects on friendships
  • 44% of Snapchat users say it helps their friendships, higher than Instagram (36%) or TikTok (24%)

• Most teens say these platforms neither hurt nor help their mental health, but parents are more negative
  • TikTok and Snapchat users lean slightly positive on mental health impact; Instagram users are split evenly between helped and hurt
  • Parents are notably more negative – 24% say social media hurt their teen’s mental health overall vs just 8% who say it helps

• Parents also view sleep and productivity impacts negatively BUT are more positive on impact to friendship
  • 41% say social media hurts their teen’s sleep; 38% say it hurts productivity
  • Friendships are the one bright spot…parents are more likely to say social media helps friendships than hurts them (22% vs. 13%)

Teens Largely Feel Good About Social Media, but Harassment Does Persist, Particularly on Snapchat

• Most teens say these platforms have no impact on how they feel about themselves
  • 57–62% across all three platforms say what they see makes no difference to their self-esteem
  • When there is an effect, it leans positive (for example, 15% of TikTok users say it makes them feel better vs only 3% who say worse)

• Overall experiences are overwhelmingly positive across all three platforms
  • Roughly 7 in 10 teens on each platform describe their experience as mostly positive
  • Just 3% say mostly negative on each platform; The rest describe an equal mix of positive and negative

• Snapchat users report the highest rates of personal harassment experiences
  • 27% of Snapchat users have experienced at least one of three harassment types (offensive name, rumor, physical threat)
  • These shares drop to about 19% on Instagram and 18% on TikTok
  • Being called an offensive name is the most common form of harassment on all three platforms

Instagram Feels Safest to Parents While TikTok Drives the Most Unease, Especially Among Parents Of Non-Users

• Parents are most comfortable with Instagram and least with TikTok
  • 59% of parents whose teens use Instagram say they are comfortable with it, vs ~51% for TikTok and ~47% for Snapchat
  • ~3 in 10 parents are uncomfortable with their teen’s TikTok or Snapchat use and that figure drops to 16% for Instagram

• Discomfort is even higher among parents whose teens don’t currently use these platforms
  • 73% of parents whose teens don’t use TikTok say they’d be uncomfortable if they did, which is the highest hypothetical discomfort of any platform
  • Smaller but still majority shares say the same for Snapchat (65%) and Instagram (52%)

After months of speculation around Globalstar’s fate that started back in October when press reports circulated that the Co was exploring a potential sale with suitors including SpaceX, Amazon was victorious and officially announced this week that it would be acquiring satellite operator Globalstar in a ~$11.6bn deal to help it build out its low-earth-orbit satellite network, Amazon Leo.

Under the agreement, Amazon will acquire Globalstar’s satellite operations, infrastructure, and spectrum assets, which will be integrated with its existing network. Amazon also announced plans to develop a direct-to-device satellite system that is expected to begin deployment in 2028. In parallel, Amazon announced a partnership with Apple to continue and expand satellite connectivity for iPhones and Apple Watches.

While the deal will expand Amazon’s network of ~200 satellites, even with Globalstar’s 24 additional satellites and FCC authorization to build an additional ~3,200, Amazon still has ways to go to match SpaceX’s Starlink constellation of ~10,000 satellites.

See below for the key details on the deal. (link)

-> Globalstar was up +10% on the back of the announcement following the +13% on deal speculation last week, while Amazon was up +4%; Globalstar shares have soared more than +93% since October 29, the last trading day before reports that the Co was exploring a potential sale

-> Also to note, AST SpaceMobile was hit hard in reaction, down -10.5%; Other Connectivity stocks were also down with T-Mobile and American Tower Corp falling -1% and -2.5%, respectively, while AT&T and Verizon were flat.

• Financial terms – Amazon will acquire 100% of Globalstar in an ~$11.6bn deal: $90.00/shr cash OR 0.321 Amazon shares (w/ value capped at $90/share)
  • Premium to GSAT stock price: Represents a +23.5% premium over Globalstar’s Monday closing price ($72.89) and a +36% premium to the share price before deal speculation emerged
  • Some caveats to note:
    • Only up to 40% of the deal can be paid in cash: If more shareholders choose cash, the extra amount will automatically be converted into stock for everyone proportionally
    • Total transaction consideration is also subject to a downward adjustment of $110mn max if Globalstar does not achieve certain operational milestones
    • GSAT stockholders holding ~58% of the combined voting power have already approved of the deal

• The deal is expected to close in 2027, subject to…
  • Customary regulatory approvals
  • Globalstar achieving certain HIBLEO-4 replacement satellite milestones

-> FCC Chairman Brendan Carr told CNBC that the agency will review Amazon’s Globalstar deal, but noted that they are “very open-minded” to the acq and it has the potential to make Amazon a competitor to SpaceX in direct-to-cell svs; “Ultimately, we’ll have to take a look at the paperwork and see, but it’s […]  consistent with the long-term vision that we have to make sure that the U.S. leads in this next-gen era of direct-to-cell technologies,” Carr said (link)

• Deal rationale – combines Globalstar’s spectrum, satellite assets, and D2D expertise with Amazon Leo to enable direct-to-device connectivity at scale, extending cellular coverage beyond the reach of terrestrial networks
  • Planning to partner with mobile network operators (MNOs) to extend existing cellular networks via satellite, positioning Leo as a complement to traditional telecom infrastructure
  • Targets consumer, enterprise, and gov’t use cases, including emergency communications, disaster recovery, remote operations, and IoT connectivity

• Underpins Amazon’s plan to build its own Amazon Leo direct-to-device satellite system, which it expects to start deploying in 2028
  • Will deliver voice, messaging, and data services directly to mobile devices
  • Offers “substantially higher” spectrum efficiency vs legacy systems, translating to faster speeds and better performance
  • The complete Amazon Leo network will include thousands of satellites with capacity to support hundreds of millions of endpoints

• Through the deal, Amazon adds Globalstar’s two dozen satellites to its existing network of more than 200
  • Amazon has also been working to ramp up its network by deploying about 3,200 satellites in Earth’s low orbit by 2029, with roughly half required to be in place by a ​July regulatory deadline
  • In comparison, Starlink has a 10,000-unit-strong network

Source: Reuters

• ALSO… Amazon and Apple annc’d an agreement under which Amazon will continue and expand Globalstar’s existing satellite services for iPhones and Apple Watchs
  • Currently, Globalstar partners with Apple to power satellite svs on iPhone 14 or later, as well as Apple Watch Ultra 3, allowing users to text emergency services, message friends and family, request roadside assistance, and share their location
    • As a reminder… Apple invested ~$1.5bn in Globalstar in 2024 to fund the expansion of its iPhone communication svs, in ⁠a deal that also gave Apple a 20% equity in Globalstar
  • With the new Amazon-Apple agreement, Amazon will continue to support iPhone and Apple Watch models currently using Globalstar’s existing and planned upcoming low Earth orbit satellite constellations, being manufactured by MDA Space, and collaborate with Apple on future satellite services using Amazon Leo’s expanded satellite network

-> Separately, but related from this week , Amazon intro’d its Amazon Leo Aviation Antenna, which allows commercial aircraft and regional jets to offer high-performance, in-flight internet to its passengers; The svs will provide up to one gigabit/sec of download speed and 400 megabits/sec of upload speed to all users aboard the aircraft; Amazon said that the new antenna was designed with no moving parts so that it can hold up under intense temperatures and difficult weather conditions, as well as to reduce maintenance downtime (link/link/link)

Source: Amazon

• Uber reportedly is making a big investment ($10bn) into the AV sector (link/link)
  • The Financial Times reported the money will be used for buying thousands of AVs and taking stakes in their developers, ‌breaking from its asset-light “gig economy” biz model to avoid disruption ​from robotaxis
    • As a reminder: The Co has partnered across much of the AV industry, including with Baidu, Rivian, and Lucid, and has outlined plans to launch robotaxi svs in ~28 cities by 2028
  • The $10bn commitment is reportedly broken down into two main pillars:
    • $7.5bn for fleet acquisition
    • $2.5bn in equity stakes

-> Also, this week, Lucid Group is deepening its partnership with Uber, Uber is increasing its purchase commitment to at least 35,000 (from 20,000) vehicles for their future robotaxi service. Uber is adding $200mn more investment (bringing its total to $500mn), while Public Investment Fund affiliate Ayar Third Investment Company is investing $550mn (link)

• According to Bloomberg’s Mark Gurman, Apple is actively testing at ~4 different styles of frames for its upcoming smart glasses project (link)
  • As a reminder these glasses are not augmented reality
    • They will have integrated cameras, microphones and sensors
    • They’ll be able to relay notifications from your phone, capture personal photos and videos, play music, and enable interactions with AI features like upgraded Siri and visual intelligence capabilities
  • The 4 designs for the Apple glasses currently in testing include:
    • A large rectangular frame, reminiscent of Ray-Ban Wayfarers
    • A slimmer rectangular design
    • Larger oval or circular frames
    • A smaller, more refined oval or circular option
  • Reported release date: Later this year or early in 2027, with an actual release in spring or summer 2027

• Allbirds -> NewBird AI: An interesting case study of a high-flying consumer brand that fell into oblivion and is now making a unique 110% pivot
  • Allbirds essentially is pivoting from a sustainable footwear brand to now an AI compute infrastructure company!
  • It sold the Allbirds brand and footwear assets to American Exchange Group (annc’d at the end of March)
  • And just annc’d this week that it raised $50mn in a convert to build its AI compute infrastructure
  • They are changing their name to NewBird AI
  • In just 5 years, the stock fell from a $4b+ valuation to being sold for $39mn 2 weeks ago

-> On the back of this new raise and strategic pivot announcement, the stock went from $2.49/shr to $17, though it is still a very micro-cap; It closed the week at $10.8 per share