The Semantic Shift: LinkedIn’s 360Brew AI System and the New Physics of Professional Visibility

The digital infrastructure of professional networking is navigating its most significant transformation since the inception of the social feed. For over a decade, LinkedIn’s algorithmic architecture operated as a “feature factory” – a complex, often disjointed assembly line of task-specific machine learning models designed to optimize for binary engagement signals such as clicks, likes, and connection requests. As the platform transitions into late 2025 and 2026, this legacy architecture is being systematically superseded by 360Brew, a 150-billion-parameter foundation model that alters the mechanisms of information distribution.¹

This research report provides a multi-dimensional analysis of this shift. By synthesizing primary engineering data from the official arXiv research papers and LinkedIn Engineering blogs with practitioner insights from industry leaders such as Richard van der Blom, Melonie Dodaro, Alexander Low, and Nigel Cliffe, this document serves as a guide for Heads of Marketing. The analysis shows a pivot from signal-based ranking – which incentivized “growth hacking” and volume – to semantic reasoning, a paradigm where distribution is governed by consistency of expertise, depth of discourse, and alignment of professional identity.

The following sections dissect the technical underpinnings of 360Brew, contrast its “decoder-only” architecture with legacy systems, quantify the new hierarchy of engagement signals (where “Saves” outweigh “Likes” by a factor of five), and outline strategic imperatives for B2B marketing leadership. The era of the viral hack is over. The era of semantic authority has begun.

Part I: The Architectural Revolution – From Feature Factories to Reasoning Engines

To understand the strategic landscape of 2026, you must first deconstruct the engineering evolution that caused it. The introduction of 360Brew is not an iterative update to ranking logic. It replaces the platform’s central nervous system, shifting from a deterministic calculator of user actions to a probabilistic reasoner of user intent.

1.1 The Legacy Architecture: The “Feature Factory” (Pre-2025)

Before the deployment of 360Brew, LinkedIn’s recommendation stack relied on a multi-stage pipeline described by engineers as a “feature factory”.¹ In this paradigm, distinct models were built for distinct tasks. A specific model would predict the likelihood of a user clicking “apply” on a job; a separate model would predict the likelihood of a user “liking” a feed post; another would suggest “People You May Know”.⁵

These systems were structurally inefficient and relied heavily on feature engineering – the manual creation of numerical signals to represent user behavior. Engineers would code specific inputs, such as number_of_comments_last_24h, shared_hashtag_count, or time_since_last_login.⁴ The algorithm functioned as a large linear regression equation, summing weighted scores to rank content.

1.1.1 The Vulnerability of Deterministic Signals

Reliance on explicit features created a deterministic environment that invited exploitation. Because the system looked for specific numerical thresholds (for example, “posts with >10 comments in 60 minutes get boosted”), practitioners could reverse-engineer the “rules.” This gave rise to engagement pods, “bro-etry” (one-line spacing for dwell time), and tag-baiting.² The algorithm did not understand what was being said. It understood how many people reacted to it.

1.2 The New Paradigm: 360Brew and the Unified Foundation Model

360Brew consolidates fragmented systems into a single, unified “reasoning engine”.⁵ Developed by LinkedIn’s FAIT (Foundation AI Technologies) team, 360Brew uses a decoder-only transformer architecture, similar to the large language models behind GPT-4 or Claude, but fine-tuned on proprietary LinkedIn data (the Economic Graph).¹

1.2.1 Technical Specifications: The Decoder-Only Transformer

The model has approximately 150 billion parameters.¹ Unlike earlier BERT-based encoders that primarily categorized text into embeddings, the decoder-only architecture supports generative reasoning and In-Context Learning (ICL). It does not just “score” a post. It reads the post, the author’s profile, and the viewer’s history as a continuous text sequence.¹

This shift to a textual interface matters. In the 360Brew era, the system constructs a “prompt” that verbalizes context. Instead of processing a numerical vector, the model processes a natural language prompt. As described in unofficial guides and engineering summaries, a typical prompt might resemble:

“The user is a Marketing Director who recently engaged with posts about RevOps. The candidate post discusses ‘AI in CRM Architecture.’ Predict the probability of a meaningful comment.”⁴

1.2.2 Universal Applicability Across Surfaces

A defining characteristic of 360Brew is universality. The official research paper states that the model is designed to power more than 30 predictive tasks across the platform, including the Feed, Job Recommendations, Search, and Ads.¹ This unification dissolves practical data silos. A signal generated in recruitment (for example, updating a skill) can inform ranking in the feed. For marketers, this implies brand consistency must be absolute; the system evaluates the “whole professional,” not just the “content creator”.⁵

1.3 Comparative Analysis: Legacy Models vs. 360Brew

The following table synthesizes operational differences between legacy LiRank systems and the 360Brew foundation model, drawing from engineering blogs and practitioner analysis.

1.4 The “Reasoning Engine” and Zero-Shot Capabilities

One of the most consequential capabilities of 360Brew is zero-shot reasoning.⁴ The model can assess relevance for content or job titles it has not encountered before by interpreting language. If a new job title emerges, legacy models typically struggle due to cold start (lack of historical data). 360Brew can infer hierarchy and relevance from the words themselves.¹

This reduces cold start for niche creators. A new voice discussing a highly specific technical topic can be matched to a relevant audience based on semantic depth, rather than waiting months to accumulate historical engagement.² This democratization is a double-edged sword: true experts can rise quickly, and established creators cannot rely on follower count alone.

Part II: The Death of Signals and the Rise of Semantics

The deployment of 360Brew has triggered what practitioners like Trey Ditto and Melonie Dodaro call “the death of the LinkedIn hack”.¹⁰ Strategies that relied on gaming the feature factory – engagement pods, optimized posting windows, and formatting tricks – are obsolete and often harmful.

2.1 The Shift from Metrics to Meaning

Under the previous regime, the algorithm prioritized velocity. If a post received a sudden influx of likes and comments within the “Golden Hour” (the first 60 minutes), the system inferred quality and distributed the post widely.⁴ 360Brew evaluates meaning.

Because the model reads content, it can separate substantive discourse from engagement bait. A comment section filled with “Great post!” or “Agree!” – common artifacts of engagement pods – is identified as low-entropy noise. The system looks for lexical diversity and conversational depth.¹³ If ten comments share similar phrasing or originate from a tight cluster of users who always engage with each other, the model downgrades reach and treats it as manufactured relevance.¹³

2.2 The “Profile-Content Audition”

A critical insight for 2026, highlighted by First AI Movers and Melonie Dodaro, is the concept of the profile-content audition.¹⁰ 360Brew does not evaluate a post in isolation. It performs a semantic cross-reference between the post and the author’s profile (Headline, About, and Experience).

The Expertise Mismatch Penalty: If a user’s profile describes them as a “Graphic Designer,” but they post advice on “Crypto Trading,” the semantic embedding of the post diverges sharply from the profile. The system detects dissonance and suppresses distribution, categorizing the content as low-trust or irrelevant.¹⁰

The Consistency Reward: When a user posts content that reinforces stated expertise (for example, a “RevOps Director” writing about “Salesforce Integration”), the system assigns higher confidence and unlocks broader distribution. This creates a flywheel of authority: consistent posting on a narrow topic progressively lowers the barrier to distribution.¹³

2.3 The “Lost-in-Distance” Effect

The “Lost-in-Distance” effect refers to how the model’s attention mechanism weights text sequences. Research indicates that opening sentences (the hook) are heavily weighted in semantic analysis.⁶

If the first two sentences are vague, generic, or fail to establish topic relevance, the model deprioritizes the post before a human user scrolls past it. Distribution is shaped by semantic clarity in the preamble. This pushes a journalistic writing style where the lead is never buried.¹⁵ Practitioners like Nigel Cliffe argue this requires a shift from “marketing copy” to “reporting,” where the value proposition is immediate and explicit.¹⁵

Part III: The New Physics of Distribution (2025-2026)

With architecture defined, the next step is to quantify the levers that influence visibility in the 360Brew era. Data analysis from Richard van der Blom’s “Algorithm Insights Report” (October 2025) and practitioner reports highlight a new hierarchy of engagement signals.

3.1 The Decline of Reach and the “Quiet Signals”

Organic reach has dropped by roughly 50% across the platform compared to prior years.¹⁷ This is a deliberate calibration to reduce noise and prioritize high-agency professional content. In this lower-volume environment, “quiet signals” replace public vanity metrics as the primary drivers of distribution.

3.1.1 The Dominance of “Saves” and “Dwell Time”

Saves (the super-signal): According to Richard van der Blom’s data, a save is estimated to be worth 5x to 10x the weight of a like.¹⁰ A save signals utilitarian, evergreen value – content a user intends to revisit. 360Brew interprets this as a high-quality signal and distinguishes useful content from merely entertaining content.

Dwell time: Time spent consuming content remains a critical input.²⁰ However, this is paired with “See More” expansion rates. If a user expands a post but abandons it immediately, it signals a clickbait hook and can trigger a penalty.

3.1.2 Comments as “In-Context” Training Data

Comments are no longer simple engagement boosters. They act as training data for the model’s understanding of the viewer.⁴

In-Context Learning (ICL): If a user leaves a thoughtful comment on a post about “AI Governance,” 360Brew updates session context and learns in real time that the user is interested in that topic.

Weighted comments: Comments that trigger replies (thread depth) are weighted higher than one-off comments.²¹ Alexander Low notes that the algorithm prioritizes conversations that resemble professional dialogue over broadcasting, rewarding posts that act as watering holes for niche communities.²⁰

3.2 Content Formats and Performance

Performance of content formats has shifted based on how well they serve the model’s preference for dense information.

Native video: Reach for native video has declined by over 70% in some analyses.¹³ This is not a blanket rejection of video. 360Brew is text-first. It reads transcripts, but video often suffers from lower completion and slower information density than text. Practitioners argue video should be paired with detailed text descriptions that the model can parse to index relevance.¹³

Document posts (carousels): These remain strong because they drive high dwell time and are text-rich, enabling full semantic indexing.¹⁹

Text + image: A practical standard for B2B. Specific, data-heavy text paired with a relevant visual (chart, screenshot) performs well because it satisfies both the AI’s need for text and the human need for visual interruption.¹⁹

3.3 The Hierarchy of Interactions

Based on practitioner data, the following table illustrates the estimated impact of user actions on post visibility in the 360Brew system.

Part IV: The Practitioner’s Perspective – Consensus and Nuance

To build a holistic view, it helps to synthesize insights from the practitioners whose work shapes how this shift is interpreted. They agree on macro-trends, but each brings a distinct lens.

4.1 Richard van der Blom: The Data Analyst

Van der Blom’s “Algorithm Insights Report” provides quantitative grounding for the 2026 landscape. His analysis supports the save multiplier and the reduction in video reach.¹⁷ He also describes “format multipliers,” noting that while content quality dominates, the container can still provide baseline advantage. He warns about staleness: even if evergreen content lives longer, the platform still biases toward recency in initial retrieval.²⁵

4.2 Melonie Dodaro: The Brand Strategist

Dodaro focuses on the human implications of 360Brew. Because the AI can detect low-quality, generic, AI-generated text (“slop”), the premium on authentic voice is higher than in earlier algorithm eras.¹³ Her work on lexical diversity in comments provides a tactical defense against engagement pod dynamics: if comments all sound the same, the system treats them as coordinated. She also argues creators must be patient, because it takes time for the semantic engine to trust a new topical direction.¹³

4.3 Alexander Low: The Social Seller

Low emphasizes the shift from broadcasting to participation. He introduces “micro-communities” – small clusters of highly relevant professionals interacting deeply.²² He argues Heads of Marketing should stop optimizing for impressions and instead measure conversation depth. He also highlights quiet signals – profile clicks, scrolls, and saves – as indicators of B2B purchase intent, which 360Brew is optimized to predict.²⁰

4.4 Nigel Cliffe: The Networker

Cliffe offers a metaphor for the new algorithm: the lunch table. In 2026, you are judged by the company you keep. If you sit at a table (comment thread) with serious professionals discussing supply chain logistics, the algorithm associates you with that table. If you sit with engagement-baiters, you are categorized as noise.¹⁵ He also distinguishes between posts and articles: posts generate short-term reach, while articles can be indexed more effectively for long-term semantic search visibility.²³

4.5 Sarah Burgess: The Career Coach

Burgess focuses on recruitment implications. She notes that many recruiters use the “Show Current” filter, which can reduce visibility for unemployed candidates unless they structure “Current Experience” strategically.²⁶ She also argues that a candidate’s activity (comments, likes) acts as a living resume. 360Brew uses that activity to match candidates to roles even when profile keywords do not exactly match, because the model interprets semantic context of engagement.⁴

4.6 Forbes (Jodie Cook): The Essentialist

Writing for Forbes, Jodie Cook frames a practical strategy as “doubling down.” She argues the algorithm rewards depth over breadth. Instead of spreading thin across multiple platforms or topics, an essentialist approach – mastering one platform (LinkedIn) and one topic – can compound because it feeds the AI consistent, high-confidence signals.²⁷

Part V: Strategic Imperatives for Heads of Marketing

For marketing leadership, 360Brew requires a shift from “social media strategy” to “semantic authority strategy.” The goal is no longer virality. The goal is to dominate the semantic clusters relevant to your product and category.

5.1 Unify the Data: Breaking the Silos

360Brew operates as a unified model across ads, the organic feed, and recruitment.⁸ This means a company’s employer brand (hiring activity) and marketing brand (content activity) are semantically linked.

Strategic action: Marketing and HR/Talent Acquisition must align narratives. If marketing posts emphasize “innovation in AI,” while job descriptions rely on outdated keywords, semantic dissonance can weaken overall entity authority for the company page. First AI Movers argues that unified systems let machine learning interpret patterns across the customer journey, creating a single narrative for the algorithm to learn.⁹

5.2 The “Topic Pillar” Governance Model

To build authority, brands and executive leaders must discipline output. The “spaghetti against the wall” approach – testing random topics to see what sticks – is now penalized.

The 90-day rule: It can take roughly 90 days of consistent posting on specific topics for 360Brew to categorize a profile’s expertise.¹³

Strategic action: Define 3-4 core topic pillars for each executive spokesperson. Ensure 80%+ of content stays within these pillars.¹¹ Off-topic deviation can dilute the semantic signal and introduce noise that lowers confidence.

5.3 Transitioning from Broadcasting to Participation

The 2026 algorithm rewards professional participation over publishing volume.¹⁵ As the lunch table metaphor suggests, users are categorized by interaction clusters.

Strategic action: Shift resources from content production to strategic commenting. Deep engagement on posts from high-authority figures can transfer semantic authority. If a CEO engages meaningfully with leaders in supply chain logistics, the model reinforces association with that topic.¹⁵

The micro-community strategy: Create high-density engagement within a niche rather than chasing broad reach. Fifty likes from relevant decision-makers can matter more to ranking than thousands of low-relevance likes, because the former strengthens the semantic cluster.²²

5.4 Content Engineering: The Utility Mandate

Content must be engineered for savability. Motivational quotes and “Happy Monday” posts are dead weight in the 360Brew calculation.

Strategic action: Mandate utility. Frameworks, detailed how-to guides, proprietary data analysis, and contrarian perspectives backed by evidence are formats that generate saves and dwell time.¹⁶

Specificity over generality: Use exact numbers, dates, and names. “We improved ROI” is weak for both humans and the model. Specific claims and concrete workflows are easier to classify, evaluate, and retrieve.¹⁶

Part VI: The Human Element – Personal Branding and Recruitment in the Semantic Era

For individual professionals and job seekers, 360Brew creates a more meritocratic opportunity: if you articulate expertise clearly, the model can find your audience regardless of follower count. It also demands higher standards of profile hygiene.

6.1 Profile Optimization as “AI SEO”

Your LinkedIn profile is no longer a static resume. It functions as the source code for semantic identity.

Headline and About: These act as primary context prompts. They should include specific keywords and entity names (software, methodologies, certifications) that define your niche.²

Skills section: While the model reads unstructured text, Skills provide structured data that corroborates the narrative of the About section. Alignment increases the confidence score assigned to expertise.⁴

6.2 The “Audition” for Job Seekers

Recruiters rely on search, and 360Brew powers that search.

Current role bias: Burgess notes that recruiters often filter by current job title. If you are unemployed, leaving “Current” blank can create a semantic gap. Job seekers can structure status (for example, “Career Break” or “Freelance Consultant”) to maintain relevance. Treating it as empty is a semantic error.²⁶

Activity as resume: The model watches what you read and how you comment. Intelligent engagement on industry posts signals competence. In-context learning can surface your profile even when you do not match exact Boolean keyword searches, because the model infers conceptual fit.⁴

6.3 The “90-Day Repair Protocol”

For professionals whose reach has flatlined (the “zombie account” pattern), a reset can be necessary.

Protocol:

• Stop posting low-value content immediately. Avoid engagement bait and off-topic posts.
• Rewrite the profile to be specific (narrow to 2-3 pillars).
• Spend 2-3 weeks commenting on high-relevance posts to retrain the model on your semantic location.
• Resume posting only high-utility, save-worthy content 1-2 times per week.¹⁰

Part VII: Future Outlook (2026 and Beyond)

Looking through 2026 into 2027, the trajectory of 360Brew suggests several trends marketing leaders should anticipate.

7.1 Agentic AI and the “Personalized Web”

360Brew is a precursor to more agentic AI experiences. The platform may evolve toward a state where users do not browse a feed. Instead, an AI agent curates: “Show me the best posts about supply chain this week.” Optimization shifts from feed optimization to agent optimization (AIO) – ensuring your content is cited as a definitive source.⁹

7.2 The End of “Format Hacking”

Some analyses argue the model is becoming more format-agnostic over time.¹³ The era of “PDFs work best this month, video works best next month” is weakening. The system increasingly prioritizes information density over container. Marketers should choose the format that best conveys the information, not the format that once gamed distribution.

7.3 Semantic Search as the Primary Discovery Mechanism

LinkedIn is positioning itself to compete with Google for B2B discovery. 360Brew’s ability to interpret natural language questions means that “SEO” on LinkedIn can converge with thought leadership. Users search for answers, and the platform serves posts that answer directly.¹⁵ This “answer engine” dynamic rewards content structured as direct solutions to complex problems.

The Era of Substance over Signal

The transition to 360Brew marks a maturity step for LinkedIn as a content platform. The feature factory era – characterized by gamification of metrics and pursuit of vanity – is being replaced by a reasoning engine that rewards semantic consistency, professional authority, and utilitarian value.

For Heads of Marketing, the mandate is clear: stop counting clicks and start engineering meaning.

The winning strategy for 2026 is not a list of hacks. It is rigorous alignment of brand identity, production of reference-grade content, and cultivation of genuine engagement within specific professional communities. The algorithm now reads like a human. To win the algorithm, write for the human expert.

Summary of Key Actions for CMOs:

• Audit profiles: Ensure executive profiles semantically match content output.
• Define pillars: Restrict content to 3-4 topics to build topic authority.
• Prioritize saves: Optimize for utility and evergreen value (high dwell time).
• Unify data: Connect marketing, sales, and talent narratives into one coherent entity for the AI.
• Engage deeply: Shift KPIs from impressions to conversation depth and saves.

The machine is listening. Say something worth hearing.

---

Sources:

1. [2501.16450] 360Brew: A Decoder-only Foundation Model for Personalized Ranking and Recommendation – arXiv, accessed on January 25, 2026, https://arxiv.org/abs/2501.16450
2. LinkedIn 360brew algorithm : Insight – Leaders Social, accessed on January 25, 2026, https://www.leaders.social/linkedin-advanced-coaching/linkedin-360brew-visibility-engagement-reach
3. 360Brew : A Decoder-only Foundation Model for Personalized Ranking and Recommendation – ar5iv, accessed on January 25, 2026, https://ar5iv.labs.arxiv.org/html/2501.16450
4. The Unofficial LinkedIn Algorithm Guide, Fall 2025 Edition – Trust Insights, accessed on January 25, 2026, https://www.trustinsights.ai/wp-content/uploads/2025/09/the_unofficial_linkedin_algorithm_guide_fall_2025_edition.pdf
5. 360Brew and the LinkedIn Algorithm: What We Know, What’s Confirmed, and What Remains Unclear? – The Linked Blog, accessed on January 25, 2026, https://thelinkedblog.com/2025/360brew-linkedin-algorithm-new-update-3619/
6. How LinkedIn’s 360Brew Algorithm is Redefining What Works on The Platform, accessed on January 25, 2026, https://topdogsocialmedia.com/how-linkedins-360brew-algorithm-is-redefining-what-works-on-the-platform/
7. Understanding LinkedIn’s 360 BREW: Why What You Post & What, accessed on January 25, 2026, https://wendyshore.co/blog/2025/11/1/linkedin-algorithm-linkedin-brew-360/
8. How LinkedIn’s new AI Model, 360Brew, signals the Future of Marketing, accessed on January 25, 2026, https://cloudanalysts.com/blog/linkedin-360brew-ai-model-future-marketing/
9. LinkedIn’s AI Shift Reveals How Platforms Now Reward Substance Over Scale, accessed on January 25, 2026, https://www.firstaimovers.com/p/linkedin-360brew-semantic-ai-visibility-strategy-2026
10. The Death of the LinkedIn Hack: How Founders Are Re-Claiming Thought Leadership in 2026 | by Trey Ditto, accessed on January 25, 2026, https://treyditto.medium.com/the-death-of-the-linkedin-hack-how-founders-are-re-claiming-thought-leadership-in-2026-420eda94e5e0
11. Unlocking LinkedIn’s 360Brew: The AI Powering Smarter Recommendations – Medium, accessed on January 25, 2026, https://medium.com/@careertips101/unlocking-linkedins-360brew-the-ai-powering-smarter-recommendations-406cec604038
12. LinkedIn’s New Brain: What 360Brew Means for Lawyers and Law Firm Marketers, accessed on January 25, 2026, https://www.tbdmarketing.co.uk/linkedins-new-brain-what-360brew-means-for-lawyers-and-law-firm-marketers/
13. LinkedIn 360Brew: What Actually Changed – AuthoredUp, accessed on January 25, 2026, https://authoredup.com/blog/linkedin-360brew
14. LinkedIn algorithm secrets, what works best on LinkedIn, accessed on January 25, 2026, https://www.thinklikeapublisher.com/linkedin-algorithm-secrets/
15. The LinkedIn Algorithm Change You Need to Know in 2026 – Converse Digital, accessed on January 25, 2026, https://conversedigital.com/linkedin-networking/linkedin-algorithm-change-2026
16. LinkedIn’s 2026 Algorithm: The Engagement Bait Era Is Finally Over – Future – Forem, accessed on January 25, 2026, https://future.forem.com/synergistdigitalmedia/linkedins-2026-algorithm-the-engagement-bait-era-is-finally-over-68
17. LinkedIn Algorithm Reset Q4 2025: Why Views Dropped and How to Fix It – Propel Growth, accessed on January 25, 2026, https://www.learning.propelgrowth.com/blog/linkedin-algorithm-reset-q4-2025-why-views-dropped-and-how-to-fix-it
18. LinkedIn Best Practice for 2025 – One to Three Marketing Solutions, accessed on January 25, 2026, https://www.onetothree.co.uk/linkedin-best-practice-for-2025/
19. Social Media Content Marketing: Top Algorithm Trends Your Team Needs to Know – The Bliss Group, accessed on January 25, 2026, https://www.theblissgrp.com/social-media-content-marketing-algorithm/
20. LinkedIn Algorithm Changes Explained | Commentify, accessed on January 25, 2026, https://blog.commentify.co/linkedin-algorithm-changes/
21. How the LinkedIn Algorithm Works (2026 Update) – SourceGeek, accessed on January 25, 2026, https://www.sourcegeek.com/en/news/how-the-linkedin-algorithm-works-2026-update
22. Master LinkedIn’s 360Brew: The New Visibility Engine for 2026 – YouTube, accessed on January 25, 2026, https://www.youtube.com/watch?v=Qx-hy_gIF7I
23. Why I’m Writing LinkedIn Articles Again! | Value Exchange, accessed on January 25, 2026, https://www.linkedintraining.co.uk/strategy/linkedin-articles-newsletters/
24. LinkedIn October Algorithm Report 2025 | PDF | Artificial Intelligence – Scribd, accessed on January 25, 2026, https://www.scribd.com/document/946320519/LinkedIn-October-Algorithm-Report-2025
25. Mastering the LinkedIn Algorithm in 2025/26 – – The Lone Recruiter, accessed on January 25, 2026, https://thelonerecruiter.com/wp-content/uploads/2025/10/Mastering-the-LinkedIn-Algorithm-in-202526-.pdf
26. Sarah Burgess, Author at Koru Associates, accessed on January 25, 2026, https://www.koru-associates.co.uk/author/sarah/
27. Doubling down – Jodie Cook, accessed on January 25, 2026, https://www.jodiecook.com/doubling-down/
28. Will Donnelly – Forbes, accessed on January 25, 2026, https://www.forbes.com/profile/will-donnelly/