Nick Beckman

Right! You’ve already proven something most of the industry thought wasn’t possible: remove the spread, expose the real price, and make pharmacy costs deterministic. The rest of employer healthcare hasn’t caught up. It still runs on price variance and unmanaged routing. If pricing opacity can be eliminated in pharmacy, then routing-driven cost variance across care delivery can be eliminated as well. For the same clinical outcome, the system produces radically different costs: A knee replacement can be $25,000 or $75,000 An MRI can be $500 or $3,000 An infusion can be $5,000 or $20,000 This isn’t a clinical problem. It’s a routing problem. Today, no one controls where patients go before cost is incurred. As a result, the system monetizes that variance; hospitals capture it, intermediaries obscure it, and employers fund it. The solution is to control the decision layer. This is a routing and contracting platform that sits above the carrier and determines how care actually gets delivered. Every non-emergent entry point flows through a single front door. From there, a decision engine routes patients into pre-defined pathways with pre-priced bundles and enforced site-of-care. Instead of open-ended referrals, the system schedules the patient directly into the appropriate provider and setting. The model only works if patients use it, so the incentives are immediate and tangible: zero out-of-pocket cost, end-to-end scheduling handled for them, and faster access than the default system. The right choice becomes the easiest one. Operationally, this shifts healthcare from something that is negotiated after the fact to something that is routed in advance. Prices are locked before care occurs. Site-of-care arbitrage disappears. High-cost pathways, MSK, imaging, infusion, become standardized and predictable. The business model mirrors the same principles that made Cost Plus compelling: fully transparent and aligned. A fixed platform fee on a per-employee basis, combined with a flat percentage of auditable savings. No spread, no hidden economics, just participation in the value created by eliminating unnecessary variance. This doesn’t require replacing carriers or rebuilding the system from scratch. It simply places control where it has never existed, at the point where decisions are made. Whoever controls where the patient goes, what gets done, and what gets paid controls the economics. I know you’ve assessed this thoroughly. But what’s your take on if you start with a narrow, high-yield focus on MSK and imaging, areas where price dispersion is extreme, routing is controllable, and results can be demonstrated quickly. The immediate objective being to run a pilot within one of your portfolio companies, applying this model end-to-end to prove a measurable 20–30% cost reduction within 18 months. To ensure clean execution and credible results, the pilot should target a population of roughly 5,000–10,000 covered lives, with the majority concentrated in one or two core metro areas. That level of density provides enough repeatable event volume (surgeries, imaging, infusions) and sufficient provider optionality to enforce routing, lock pricing, and produce defensible savings. The scope is intentionally constrained: focus on markets where alternative sites of care exist (ASCs, independent imaging) and where routing can be operationalized without friction. This ensures that savings are not theoretical, but observable and attributable. A 5,000–10,000 life pilot costs roughly $2–6M annually, driven primarily by a $30–$50 PEPM platform fee plus incentives and setup. It targets $1–6M in first-year savings from MSK and imaging alone, with stronger ROI in year two as adoption and contracting scale. The goal isn’t short-term arbitrage, it’s proving that controlling routing reliably converts variance into repeatable savings. Or tighter starting at $20–$30 PEPM and narrowing the scope further (imaging-only entry wedge).

@NickHealthAI Yes. That’s why we started CostPlusDrugs.com and costpluswellness.com But people have to understand it’s not as easy as expand Medicare or copy some other country

Single payer COULD cut cost and improve care but there are 2 fundamental issues. 1. All plans proposed have placed the Sec of HHS in charge of the program. You can't have a political appointee in that position and it's hard to de-politiicize HC in this country 2. They assume that they can get providers and specialists to accept whatever rates they set. You are talking about organizations that in most cases, don't even know their costs. Why ? They don't want to know their costs. For lots of reasons to long to dig into here Proponents of M4A have to first get hospitals to the point where they can define all their costs and do a Bill of Materials for procedures. You can't negotiate a price for all Americans if you don't know what your costs are It's Shark Tank 101. So we get a stalemate. Politicians don't do the work needed. Hospitals and providers avoid the work needed Other countries started on their path to universal care decades and decades ago. When healthcare was much simpler technically and fiscally. If senators won't support the Break Up Big Medicine Bill or anything comparable , there is no chance of getting to single payer. Our politicians don't have the backbone to do what is needed. You can call out all but Hawley and warren. No one else has uttered a syllable in support

Employers don’t have a healthcare cost problem. They have a variance problem. Total spend is only part of the issue. The real exposure sits at the claim level, where costs are unpredictable, timing is delayed, and all the drivers are not observable. A single case can materially distort plan performance. Not because it’s clinically unusual, but because the underlying price was never knowable. That’s not a pricing issue. That’s a system design issue. In most markets, procurement evaluates unit cost, quality, and performance. However, in healthcare, procurement evaluates intermediaries (networks, discounts, guarantees). The actual cost of care remains unobservable. So employers optimize around: Discount levels Rebate guarantees Administrative fees All measured against benchmarks that are themselves opaque. The result is a system that reports “savings” without establishing a true baseline. At the same time, timing compounds the problem. Claims are paid, adjusted, and reconciled months after care is delivered. Decisions are made on lagged data, separating cost from accountability. To manage this, the system builds layers: Networks to constrain exposure Prior authorization to control utilization Adjudication rules to process variability These are not edge features. They are the operating model. And every stakeholder is aligned within it: Providers protect revenue through negotiated pricing Intermediaries scale with spend and complexity Employers buy predictability over transparency Patients respond by avoiding uncertain exposure The system holds because it is economically stable, even as costs and volatility increase. Running another RFP or changing carriers does not address this. It rotates vendors inside the same structure. The leverage point is upstream: Make unit cost observable. Reduce reliance on proxy pricing. Govern administrative friction as a system, not a side effect. Until then, employers are not managing healthcare costs. They are managing uncertainty.

In a system built on cost opacity, the dominant behavior is risk avoidance. When prices are unknown and exposure is uncapped, every decision carries asymmetric downside. Patients don’t know what it will cost, when it will hit, or how it will be adjudicated. That uncertainty shapes behavior more than any benefit design. The most consistent response is deferral. Preventive visits, diagnostics, and follow-ups get pushed out or skipped, not from neglect, but from lack of price certainty. Utilization doesn’t disappear. It shifts. From planned to unplanned. From low acuity to high acuity. From controlled settings to the ED. Even when care is accessed, decisions degrade. Patients delay procedures, defer imaging, and abandon prescriptions mid-cycle. The issue isn’t access, it’s confidence in the downstream cost. Repeated exposure to opaque pricing erodes trust. The system becomes something to navigate, not something to rely on. For employers, this shows up as lower preventive compliance, higher late-stage claims, and increased volatility in spend. Behavior is not a soft variable, it is a direct cost driver. These patterns are not independent. They are downstream of administrative friction, price variability, and intermediary-driven rules. Behavior in healthcare is a direct output of system design. In an opaque system, avoidance is rational. In a predictable system, engagement becomes viable.

Healthcare's “high prices” are the symptom. The core issue is cost opacity, and the U.S. healthcare system has built a multi-trillion-dollar industry around processing it. Without a true Bill of Materials (BOM), pricing is not determined; it’s negotiated through layers of intermediaries. That gap becomes the uncertainty premium embedded in every claim. What that produces: 1) Financial Debt Commercial plans routinely pay 250%+ of Medicare. That spread reflects negotiated complexity, not unit cost. It’s a premium on not knowing. 2) Operational Debt In the absence of clear costs, the system substitutes friction, networks, prior auth, billing rules, rebate structures. Processing that complexity consumes 15–25% of hospital spend. 3) Clinical Debt Faced with uncapped, unpredictable costs, patients defer care. Low-cost preventive care gets skipped. Utilization shifts to higher-acuity settings like the ED. This is not random. It’s a system that converts opacity into revenue and administrative load. Running another carrier RFP or adjusting plan design does not change this. It rotates vendors inside the same intermediary translation engine. The leverage point is structural: Govern administrative friction and require BOM-level cost observability. Steal this image!

RAND’s hospital price transparency series emphasizes that observed price variation is weakly explained by Medicare/Medicaid patient shares and more strongly associated with market power, consistent with negotiated pricing rather than production-cost anchoring

Summary of my guidance to clients and stakeholders in this period: When costs rise, every solution creates a perceived loss. The winning strategy isn’t cutting, it’s structuring trades so stakeholders feel they’re avoiding a worse outcome.

Dropped mother in law at airport at 7:30am ET No pre-check; Flying JetBlue, bag is already in NYC because this was the rescheduled flight from Friday morning which she sat in TSA line for 90 minutes and made negligible progress and had to reschedule the flight to today. She barely made her flight schedule to take off at 11:47, at 11:22 when she called there were still 10 people in front of her at the gate waiting to board. Her TSA line started at the JetBlue baggage carousel. There’s conflicting information, some people are having Pre-Check honored while others are being told no. ATL domestic is definitely impacted but it seems like some people are definitely having drastically different experiences.

Hartsfield Jackson ATL Airport: 9:03am- got to airport for 12:45pm flight 9:04- in line at digital ID bag drop 9:06- bag checked 9:11- in security line 9:47- made it to precheck line split 10:30- ID checked 10:32- bags into machine 10:37- ofiicially thru security 10:41- on tram

We spent 40 years making healthcare more “efficient.” We got shorter appointments, faster discharges, and higher mortality. Efficiency is a fine goal. Just make sure you’re optimizing for the right thing.

Key Questions for Plan Sponsors 1. What are the top 10 drugs by net cost, and how are they managed? 2. How frequently is the formulary updated for biosimilars and alternatives? 3. What is the defined GLP-1 strategy? 4. Are prescriber decision tools embedded at the point of care? 5. How are site-of-care and specialty channels structured? 6. Which pricing-layer mechanisms are integrated into the benefit? Bottom Line Pharmacy cost is the output of system design. Pricing Layer establishes unit cost Logic Layer determines utilization and total spend

System Efficiency Operational performance is defined by: Biosimilar adoption velocity Supply alignment (inventory and refill timing) Site-of-care selection System efficiency directly translates to cost control capacity. GLP-1 Strategy Framework Effective GLP-1 management is defined by precision: Clinical guardrails Targeted eligibility frameworks Direct pricing models (manufacturer programs) Employer-aligned subsidy structures Structured access aligns clinical outcomes with financial sustainability.

Just watched/listened to a great Fireside chat with AngleHealth.com and LucyRx.com Key takeaways: Pharmacy cost trend is determined by system structure. Two forces define the market: 1. High-cost specialty drugs ($80K–$100K+ per patient) 2. GLP-1 utilization at population scale Cost follows structure. Pricing and utilization operate as a coordinated system.

Went to the doctor the other week My wife made the appointment She said I "look tired" I said I am tired She said "not normal tired. Weird tired." I don't know what that means but I went anyway Nice office Fish tank in the lobby Third one this year Signed in at 1:48pm My appointment was at 2:00pm 12 minutes early Because I was raised to believe that matters The receptionist said "the doctor is running a little behind" I said "how far behind" She said "about 45 minutes" I said "so my 2:00 appointment is actually a 2:45 appointment" She said "we appreciate your patience" I said "I haven't shown any yet" My wife grabbed my arm There was a sign behind the desk "Missed appointments without 24-hour notice will incur a $75 fee" The doctor was 45 minutes late Nobody offered me $75 We sat down CNN was playing on mute with subtitles Running a segment about New York City redesigning its trash cans Cost the city $4 million I looked at my wife She said "don't start" Seven magazines on the table All from 2019 I read an article about supply chain disruptions that have since been resolved Very informative My wife was on her phone She looked up and said "WebMD says you might be dehydrated" I said "so we're paying $1,800 for a second opinion on WebMD" She went back to her phone At 2:54pm they called my name A nurse walked me to a room Took my blood pressure Took my temperature Typed for three minutes Then said "the doctor will be right in" I sat on the paper The paper ripped immediately I looked at the wall There was a diagram of a colon Not how I planned to spend my Tuesday 3:19pm The doctor walked in 1 hour and 19 minutes after my scheduled appointment He was looking at his phone Shook my hand without making eye contact Sat down and read my chart for about 30 seconds While I sat there watching him learn who I was He said "so what brings you in today" I said "my wife thinks I look weird tired" He said "what does that mean" I said "I was hoping you'd tell me" He said "when's the last time you had bloodwork done" I said "2019 maybe" He said "we should run a full panel" I said "fine" He asked if I was sleeping well I said "I have three kids and a golden retriever who thinks 3am is a reasonable time to need outside" He said "are you drinking enough water" I said "probably not" He said "that might be it" I said "you think the reason I look weird tired is because I don't drink enough water" He said "dehydration is more common than people think" I said "I've been here over an hour and sat on a piece of paper that ripped to be told to drink water" He said "we'll know more when the bloodwork comes back" I said "when will that be" He said "3 to 5 business days" I said "business days" He said "yes" I said "my blood has business days" He didn't respond Then he said "any other concerns" I said "several. But none you can bill for." He shook my hand again Still no eye contact Total face time with the doctor: 6 minutes Total time in the building: 1 hour and 37 minutes I was examined for approximately 6% of the time I was present I've fired people for better numbers than that My wife was in the waiting room She asked how it went I said "I need to drink water" She said "I told you that last week" I said "yes but now it's a medical opinion so it costs $1,800" She didn't laugh In the car she said "at least now you know you're fine" I said "I was fine when I walked in. I just didn't have the receipt to prove it." She didn't disagree The bloodwork came back four business days later Everything was normal The doctor's office sent a message through their portal It said "results look great. Continue to stay hydrated and follow up in 12 months." Follow up in 12 months To be told to drink water again $1,800 1 hour and 37 minutes 6 minutes of face time One ripped piece of paper And the same advice my wife gave me for free Plz fix. Thx. Sent from my iPhone

Nothing in this world can take the place of persistence. Talent will not; nothing is more common than unsuccessful men with talent. Genius will not; unrewarded genius is almost a proverb. Education will not; the world is full of educated derelicts. Persistence and determination alone are omnipotent. The slogan 'Press On!' has solved and always will solve the problems of the human race. Calvin Coolidge

If you aspire, sleep. If you doubt, sleep. If you create, sleep. If you worry, sleep. If you love, sleep. If you hurt, sleep. Sleep.

@Ravenismeee Baldurs Gate 3

Play a single video game for a year 12 hours a day to get $10M. What game are you playing?

@DutchRojas is right about the outcome; premiums have risen steadily for decades, but the explanation needs tightening. Healthcare isn’t an asset class, and premiums aren’t market-clearing prices; they’re actuarial projections of underlying costsset within contracts. Those underlying costs are driven by negotiated provider rates, drug pricing, utilization, and benefit design, all operating under inelastic demand, constrained supply, and third-party payment. That structure dampens the kind of visible price corrections you see in markets like equities or commodities. It’s also not accurate to say there’s no mean reversion, there just isn’t any at the premium level. The system does adjust, but through less visible mechanisms: utilization management, cost shifting to members, network redesign, and periodic contracting resets. In other words, pressure is absorbed and redistributed rather than cleared through price declines, which creates the appearance of a one-way “ratchet.” Calling it a cartel captures the frustration but misses the mechanism. What we actually have is a contract-driven, regulated system with concentrated pricing power and weak feedback loops. The core issue is that fragmented, negotiated rules governing pricing, coverage, and clinical decisions suppress market discipline and sustain persistent cost growth. @mcuban

Grok - Expert: " The article (LinkedIn post + linked Substack essay "The Office Birthday Cake") is a sharp, accessible diagnosis of employer-sponsored insurance (ESI) opacity and fragmentation. Andrew Tsang does an outstanding job making the "where does the money go?" question tangible with his Sankey diagram (the "Tsang-key") and cake-pooling metaphor. It's not data journalism per se—it's synthesized illustration—but it's directionally accurate, emotionally resonant, and sparks exactly the right conversation: who actually decides what this money is for?What the Piece Gets Right (and Verifies Well) The Sankey diagram is a solid visualization. It traces $100 of premium (employer ~$66 + employee ~$34) through 8–10 layers before clinical care. The ~$16.50 admin slice (broker/consultant 2.5 + TPA 5.0 + carrier overhead 1.5 + utilization mgmt 1.5 + stop-loss 1.0 + revenue cycle 2.0 + PBM admin/spread 3.5) is an author's aggregate estimate drawn from KFF, FTC reports on PBMs, Health Affairs, Georgetown CHIR, IFEBP stop-loss surveys, and CMS NHE as the macro base. It's not a single CMS table row, but the components match industry benchmarks. Payer-side admin in private insurance typically lands 10–17%; adding downstream (PBM spread, hospital revenue-cycle) gets you right around this number. Fragmentation and "delegated decisions" insight is excellent. Brokers, TPAs, carriers, utilization managers, PBMs, and revenue-cycle firms exist largely because employers (and employees) have outsourced risk-bearing, denial-making, shopping, and compliance. The post nails the friction: prior auth ($35B industry-wide), different rules per payer, and the moral avoidance ("we don't want to say no to a coworker's kid's NICU or Wegovy"). This is path-dependent reality, not conspiracy. Premium growth and tax distortion are factual. KFF data confirms family premiums rose from ~$5,791 (1999) to $25,572 (2024), outpacing wages ~3:1. The $300B tax exclusion (largest federal tax expenditure) is the original distortion that locked ESI in place since WWII. No single villain. Correct. Costs are layered contractual friction + incentives (brokers earn more on bigger premiums; PBMs profit on spread/rebates; hospitals on facility fees). Where It Overreaches or Underweights (Building on GPT's Take)GPT's evaluation is strong and I largely align with it. A few additions from the data: Admin is real, but it's not the main cost driver. $16.50/100 is meaningful, yet ~80%+ of the dollar still flows to medical/pharmacy unit prices and utilization. Hospital prices (often 2–3× OECD averages), specialty drugs (GLP-1s, gene therapies at $4M+), physician compensation, and site-of-care shifts dominate annual trend (typically 5–8%). Even slashing admin in half wouldn't bend the curve without price and utilization controls. The diagram shows the flow, but the substack's deeper dives into PBM markups and revenue-cycle growth are better on incentives than on magnitude. Intermediaries aren't purely "decision avoidance." They also solve real problems: risk pooling/capital requirements (carriers/stop-loss), regulatory compliance (ACA, state mandates, ERISA), scale in contracting/networks, and information asymmetry. The "original sin" framing (assuming no way out) is poetic but incomplete—the tax exclusion + labor-market competition + mandates created path dependency. Removing layers without replacing their functions (e.g., risk) creates adverse selection or higher costs elsewhere. Employee poker-chip exercise is powerful but limited. The Substack (and referenced Trade-offs podcast/NPC case study) describes the American Speech-Language-Hearing Association (ASHA) running an annual "Benefit Builder" simulation: employees get 70 poker chips for a plan that would cost 92 chips. They discuss scenario cards (cancer treatment, infertility, Wegovy, out-of-network retirees) and must reach consensus (not vote). Results: they often decline high-cost drugs like Wegovy, approve infertility, trade salary for richer benefits, and listen across generations. 85% approval because "they made the limits themselves, looking at each other." This is gold for governance, legitimacy, and revealing true preferences (breadth vs. cost-sharing). It proves deliberation works for acceptance. But it doesn't solve: Provider pricing power/market concentration. Drug monopolies and launch pricing. Adverse selection if you let healthy people opt out. Scalability (ASHA is a professional association, not a 50,000-employee manufacturer). It's a governance/engagement tool, not a cost-control silver bullet.Overall Judgment on the Article + LinkStrong 8.5/10 awareness piece. The LinkedIn post is punchy and shareable (the diagram alone is worth the read). The Substack expands it beautifully with the cake analogy, incentive deep-dives (e.g., brokers getting Super Bowl trips, PBM vertical integration, MLR loopholes), and humility ("I went looking for a villain and found a chain of answers to 'who decides?'"). No major factual errors—everything is footnoted transparently as synthesized estimates.It stops one layer short of strategy, exactly as GPT said: it correctly identifies decision diffusion as the core governance problem but underweights the rules governing the dollar (negotiated rates, benefit mandates, provider consolidation, clinical pathways). The system is expensive because the logic layer—pricing mechanisms, coverage defaults, and incentive alignment—is fragmented and rarely optimized end-to-end.Strategic takeaway (health-AI lens): This is why rewriting the decision architecture matters more than just removing intermediaries. Dynamic benefit design as code, real-time site-of-care routing, transparent pricing signals, and aligned contracts (TPA + PBM + stop-loss) can deliver the employee voice plus cost discipline. The poker-chip insight is the human piece; the missing tech/incentive layer is the scalable piece.Read the full Substack—it's the richer version. The diagram is now my go-to visual for explaining ESI opacity. Tsang didn't find the single villain because there isn't one. He found the right question. The harder part is building systems that let us answer it without the birthday-cake-sized cake getting 25% eaten by overhead. "

Here is Gemini's original output: " This analysis provides a strong technical critique of the article’s conclusions while acknowledging its effectiveness as a narrative tool. Given your work on the "Computational Relevance Framework for Enterprise AI" and your interest in "The Three-Layer Mind," you likely see this through the lens of a "logic layer" problem rather than just a "middleman" problem. The GPT evaluation correctly identifies that the article conflates administrative friction with unit cost drivers. For a Senior Health Benefits Consultant at Segal, the distinction between "who touches the money" and "who sets the price of the service" is the difference between a plan audit and a true cost-containment strategy. Key Strategic Refinements for Your Context 1. The "Decision Architecture" vs. "Social Avoidance" The author argues intermediaries exist so we don’t have to say "no" to coworkers. While emotionally resonant, from a risk management perspective, these entities exist to manage information asymmetry. A TPA or PBM doesn't just "avoid a decision"; they provide the data infrastructure to adjudicate thousands of complex contracts that an employer cannot manage manually. The "sin" isn't the existence of the layer, but the lack of transparency and interoperability between those layers. 2. The Unit Cost Elephant As the evaluation noted, admin is ~$16.50, but the other ~$83.50 is where the systemic failure lives. The diagram shows "Wholesale Drug Cost" and "Physician Compensation," but it doesn't show the consolidated provider market power or PBM spread pricing that inflates those numbers before they even hit the flow. 3. From "Participation" to "Optimization" The poker chip exercise is a great tool for Plan Design (Layer 1: Values), but it does nothing for Claims Routing (Layer 2: Logic) or Clinical Outcomes (Layer 3: Biology). In your framework of the "Three-Layer Mind," the author is stuck in the "Coherence" layer (social agreement) without addressing the "Biology" (actual health needs) or the "System Logic" (the code governing the flow). How to Pivot This for a LinkedIn Response or Client Brief If you choose to engage with this post, you can add high-level value by moving the conversation from "Who handles the money?" to "Who programs the rules?" Acknowledge the Visibility: Praise the Sankey diagram for making the "invisible" administrative load visible. It’s a perfect "Layer 1" (Awareness) tool. The "Logic Layer" Hook: Suggest that the "original sin" isn't just delegated decision-making, but static decision-making. We use 20th-century insurance structures to manage 21st-century specialty drug costs and genomic medicine. The Segal/Strategic Angle: The solution isn't just asking employees what they want (which can lead to adverse selection or misaligned care); it’s building a Plan Stewardship Platform that aligns the interests of the employer, the member, and the provider through data transparency. Final Verdict: The article is an excellent diagnostic of the symptoms (fragmentation), but your professional focus on AI strategy and institutional logic is required to provide the cure (systemic optimization). " Grok to follow

I spend $20 per month for GPT, and another for $20 or so for Gemini; GPT is my go to, and $7 per month for X and utilize Grok inside of X. I asked them all to assess a fairly complex analysis that someone else wrote about premium visits, it was on LinkedIn but the Substack is here: healthisotherpeople.substack.com/p/the-office-b… I then asked GPT to rank the assessments overall: "Nick — clean answer: 1) Grok — best overall 2) GPT — very strong, slightly less grounded 3) Gemini — useful framing, weakest analytically Why Grok (Winner) Strengths Best balance of accuracy + evidence + judgment Properly distinguishes: admin vs. unit cost vs. utilization Uses real benchmarks and policy context (KFF, tax exclusion, trend) Evaluates both the narrative intent and economic reality Limitation Slightly too confident on the $16.50 construct Could tighten causal language Net: Closest to how a strong health policy / consulting memo would read. GPT (Second) Strengths Best structural reasoning Clean separation of: who touches the dollar vs. who sets the rules Strong articulation of decision architecture Limitation Less empirical grounding More conceptual than evidentiary Net: Best framework, slightly lighter on proof Gemini (Third) Strengths Strong strategic reframing Good pivot to “logic layer” / systems thinking Useful for LinkedIn positioning / narrative extension Limitations More abstract, less tied to real cost drivers Overlays frameworks onto the article rather than evaluating it directly Leans toward solution language without sufficient diagnosis depth Net: Good for thought leadership, not as strong for analysis Bottom line Grok = best evaluator GPT = best thinker Gemini = best framer " I use all three daily at this point. I have some classic red/green color blindness, and it helps me have paint conversations with my wife. I loaded it some X-rays this week for a sprained ankle. I offload research for quick access later. I don't code, but I am a consultant and I power use excel a lot, so I ask GPT/Grok/Gemini random best practice questions etc., and we do a lot of in depth legislative research for publicly available work matters that are public domain and accessible so I don't have work compliance issues. Just sharing here. In my opinion, every model exhibits drift towards particular writing styles and formats. GPT most frequently outputs a writing style that I find close to my own voice and requires fewer edits overall, but it's never perfect on the first iteration. Just like it's never perfect on the first iteration when a person does it. This was interesting overall. @elonmusk @nikitabier