• Anthropic’s stance against the Pentagon and OpenAI’s resulting agreement is shifting the chatbot wars.
• As some ChatGPT users posted about canceling, Anthropic’s Claude over took ChatGPT to hit No. 1 on the App Store.
• OpenAI said its Pentagon agreement emphasizes human oversight of autonomous weapons and limits mass surveillance.

While OpenAI locks down Washington, Anthropic is locking down users and rocketing to the top of the App Store.

Anthropic has been sidelined in Washington following a public dispute with the Department of Defense over how its AI models would be deployed. President Donald Trump ordered federal agencies to phase out its technology.

Meanwhile, OpenAI has secured new ground, with CEO Sam Altman announcing in a Friday night post on X that it had reached an agreement with the Department of Defense to deploy AI models in its classified network.

OpenAI’s agreement has left some loyal ChatGPT users uneasy about OpenAI’s ambitions, prompting online debates about the ethical implications — and some saying they were defecting to its rival Claude.

As of 12:50 pm ET on Monday, Claude ranked No. 1 among the most downloaded free apps on Apple’s US App Store. Claude also tops the chart for top productivity apps, where the top four apps are all AI competitors: ChatGPT, Google’s Gemini, and xAI’s Grok.

A spokesperson for Anthropic said, “Every single day last week was an all-time record for Claude sign-ups.” The AI startup said its upswing dates back to its much-discussed playful Super Bowl ad, which mocked OpenAI’s decision to test ads on ChatGPT.

“Since the start of the year, free active users have increased by over 60% and daily signups have quadrupled,” the spokesperson wrote in an email to Business Insider. “Claude’s paid subscribers have also more than doubled this year across Pro and Max plans.”

Even with the increase, Anthropic has a long way to go in matching ChatGPT, which OpenAI said on February 27 boasts over 900 million weekly active users. Anthropic CEO Dario Amodei has also positioned his company to focus more on the enterprise market.

Converts have taken to social media to share screenshots documenting their switch.

Pop musician Katy Perry wrote that she was “done” on X, alongside a screenshot of Claude’s pricing page, with a red heart around the $20-per-month “Pro” plan.

Another X user, Adam Lyttle, wrote “Made the switch,” alongside a screenshot of his email inbox with a receipt from Anthropic and cancellation confirmation from OpenAI.

On Reddit’s ChatGPT subreddit, dozens of users say they’ve deleted their accounts and are urging others to do the same.

“Cancel ChatGPT” has become a common refrain online, while some users have taken a more personal tone, saying Altman’s move “crossed the line.”

The agreement hasn’t polarized all AI users, however.

In one Reddit thread, several commenters said the news does not affect their choice of AI model, arguing that Anthropic’s work with Palantir raises similar concerns. In November 2024, Anthropic, Palantir, and Amazon Web Services struck an agreement to provide US intelligence and defense agencies access to Claude models.

After Secretary of Defense Pete Hegseth said he would designate Anthropic as a “supply chain risk to national security,” Anthropic said it would “challenge any supply chain risk designation in court.”

In his Friday post, Altman said the Department of Defense had agreed with two of OpenAI’s safety principles.

“Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems,” Altman wrote on X. “The DoW agrees with these principles, reflects them in law and policy, and we put them into our agreement.”

By Saturday afternoon, OpenAI published a more detailed description of its contract with the Department of Defense, including the specific language it used surrounding the use of its models for surveillance and autonomous weapons.

On the topic of autonomous weapons, OpenAI said:

The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control, nor will it be used to assume other high-stakes decisions that require approval by a human decisionmaker under the same authorities.

On the topic of mass surveillance, OpenAI said:

The AI System shall not be used for unconstrained monitoring of U.S. persons’ private information as consistent with these authorities.

While some chatbot users suggested it’s all fair in business, war, and federal procurement, others suggested the Pentagon’s stance may have handed Anthropic a public relations win.

X user Tae Kim joked that Hegseth might need a new title: “Secretary Hegseth Chief of Claude Marketing.”

Chef and restaurateur Cesare Casella breaks down Italian cuisine scenes from movies and TV based on realism.

Casella discusses the accuracy of the meatballs from “The Godfather,” pasta sauce and thin garlic in “Goodfellas,” and capicola in “The Sopranos.” He also analyzes a dish from “Lady and the Tramp” (1955), timpano from “Big Night,” the tiramisu in “Superbad,” and carbonara from “Master of None.” He breaks down the pasta twirling technique from “Brooklyn,” cannoli from “The Bear,” and calzone from “Parks and Rec.” Casella reviews proper pizza throwing technique from “Seinfeld,” pizza-eating form in “Eat Pray Love,” lasagna-making in “Garfield: A Tale of Two Kitties,” mozzarella sandwiches from “Bicycle Thieves,” take-out salads from “Family Guy,” Italian kitchen-table dining in “Amarcord,” and trenette al pesto from “Luca.”

Casella has been cooking for more than 50 years and served as the head chef at the Michelin-starred restaurant Vipore in Italy. He then moved to the United States and worked at several fine-dining Italian restaurants in New York City such as Beppe and Salumeria Rosi. Today, Casella is head of the Department of Nourishment Arts at the Center for Discovery, a residential facility in upstate New York for the medically fragile. He oversees a team of chefs, nutritionists, therapists, farmers, and educators.

You can find Cesare on Instagram.

• In December, Paramount’s Larry and David Ellison said they had partners to help them buy WBD: Three Gulf states.
• At the time, Saudi Arabia, Abu Dhabi, and Qatar were going to chip in a collective $24 billion into the deal.
• Are those countries part of the new deal Paramount struck with WBD last week? We don’t know yet.

A few months ago, Larry and David Ellison’s Paramount said it had lined up big backers for its bid to acquire Warner Bros. Discovery: three Middle Eastern countries, which would contribute a total of $24 billion to help buy the media conglomerate.

Now, Paramount has won the rights to WBD, after Netflix walked away. Are those three Gulf states going to help finance Paramount’s new deal?

Paramount won’t say.

In government filings and on an investor call Monday, Paramount reiterated that the Ellisons and private-equity firm RedBird Capital Partners have pledged $47 billion toward the roughly $81 billion Paramount will pay to buy out WBD shareholders. The rest will be financed with debt.

But Paramount doesn’t say how much the Ellisons and RedBird intend to cough up themselves, and how much will come from other investors. A company rep pointed to the merger agreement, which lets them sell pieces of that $47 billion equity commitment — but doesn’t identify who those investors might be.

What we do know: In early December, Paramount had told WBD it intended to raise $24 billion via state-controlled funds from three countries: Saudi Arabia, which would contribute $12 billion, and Abu Dhabi and Qatar, which would kick in $7 billion each. And Paramount later said those countries would not ask for board seats or voting rights alongside their equity stakes.

So it’s entirely possible that those three countries’ funds will be in this deal, too. I’ve asked fund reps for comment; I’ve also asked Warner Bros. Discovery, which had previously flagged potential regulatory issues involving foreign investors in Paramount’s deal.

The notion of other countries owning a piece of a powerful American-run media giant, which has multiple news operations, streamers, and movie studios, seems like the kind of thing you might hear politicians, regulators, and even normal Americans talking about. Particularly given the participation of Saudi Arabia, whose government killed a Washington Post journalist in 2018, per US intelligence.

It may also explain why Paramount has never emphasized the idea of foreign investment in any of its press releases or public presentations about the deal. Or why Paramount once shot down a story saying it was rounding up Gulf investment.

But outside of a few complaints from the likes of Sen. Elizabeth Warren, the notion that Middle Eastern nations might own a meaningful stake in an important US company doesn’t seem to have registered as an issue, for now.

One reason is that Gulf state money is increasingly everywhere, including in big US companies. The Saudis, for instance, are set to be the primary owner of video game giant Electronic Arts once that deal closes this year. Abu Dhabi’s MGX fund now owns 15% of U.S. TikTok‘s operations.

And Donald Trump has made doing business in the Middle East a big part of his second-term focus: Last year, he brought a who’s who of American CEOs to a business summit in Saudi Arabia and later hosted Saudi Crown Prince Mohammed bin Salman at the White House.

So it’s very possible Gulf state money will end up in a giant American media company, too. It would be nice to know for sure.

For many enterprises, the hardest part of an IoT program is not selecting sensors, connectivity, or platforms. It is proving—year after year—that the deployment is economically sound. The conversation is shifting from “Can we fund the initial rollout?” to “Can we sustain and scale the operating model?”. This is why Total Cost of Ownership (TCO) has become the most practical decision framework for IoT: it exposes what truly drives long-term cost, risk, and value—well beyond the hardware bill of materials.

At the same time, IoT economics are changing. Device lifecycles are lengthening in industrial environments, security and compliance costs are rising, connectivity pricing is diversifying, and cloud spending is under tighter governance. As a result, many buyers are moving from CapEx-heavy project thinking to OpEx-driven product thinking—where ongoing service delivery, reliability, and change management matter more than the initial procurement.

Why IoT TCO models need a refresh in 2026

Classic IoT business cases often over-index on upfront costs (devices, installation, and initial integration) and under-estimate what happens after go-live. Three forces make that gap more expensive:

• Security and resilience are now baseline requirements: secure provisioning, key management, vulnerability monitoring, incident response, and periodic remediation need to be budgeted as recurring costs.
• Cloud and data costs are scrutinized: ingesting “everything” is rarely sustainable; storage, egress, and analytics workloads must be modeled with realistic usage patterns.
• Operational complexity is rising: multi-vendor stacks, multiple regions, device fleet growth, and regulatory constraints increase the cost of change.

A modern TCO model therefore needs to be less about “project cost” and more about “cost-to-serve a device, a site, and a business outcome over time.”

A practical IoT TCO framework

In IoT, TCO is best modeled in layers. Start with a baseline per-device/per-site cost structure, then scale it across fleet size, geography, and service levels (SLA). A robust framework includes:

1) CapEx: what you pay to deploy

• Device hardware: sensors, gateways, modules, enclosures, power systems, certifications, spares.
• Installation and commissioning: site surveys, labor, travel, safety constraints, calibration, acceptance testing.
• Initial integration: connectors to enterprise systems (ERP/MES/CMMS), data mapping, dashboards, identity, initial automation rules.
• One-time security setup: provisioning tooling, PKI bootstrap, secure element options, manufacturing or onboarding processes.

2) OpEx: what you pay to operate

• Connectivity recurring fees: SIM/eSIM subscriptions, private network costs, LPWAN subscriptions, roaming, APNs, VPNs.
• Cloud/platform consumption: message ingestion, device registry, digital twins, rules engines, storage, analytics, alerting, logs, egress.
• Device management and firmware operations: monitoring, remote diagnostics, configuration changes, OTA updates, rollout validation, rollback procedures.
• Security operations: certificate rotation, vulnerability management, pen tests, monitoring, SOC integration, incident response, compliance reporting.
• Field operations and support: truck rolls, replacements, RMAs, spare management, tiered support, NOC processes.
• Data operations: data quality checks, model maintenance, labeling (if AI), integration maintenance, API lifecycle management.

3) Risk and cost of change: the hidden multiplier

Many IoT deployments fail to scale because the “cost of change” is not modeled. Every change—new site, new device variant, new security requirement, new regulation, supplier disruption—creates engineering, validation, and operational work. Treat this as a budget line, not as a surprise:

• Change requests and roadmap delivery: feature evolution, new dashboards, new KPIs, new workflows.
• Vendor transitions: module swaps, carrier changes, cloud migrations, platform re-architecture.
• Regulatory updates: privacy/data retention, critical infrastructure constraints, sector compliance.

Core modeling units: per device, per site, per outcome

To keep models actionable, avoid a single monolithic spreadsheet. Instead, build three units and connect them:

• Cost per device-year: a normalized number that includes connectivity, platform, device management, and support (plus allocated security operations).
• Cost per site-year: site-specific fixed costs such as gateway infrastructure, on-prem edge compute, private network components, site visits, and local compliance requirements.
• Cost per outcome: the cost to deliver a business KPI (e.g., cost per avoided downtime hour, cost per tracked asset, cost per compliance report).

This structure helps decision makers compare architectures (direct-to-cloud vs gateway/edge, single-carrier vs multi-carrier, centralized vs regional data) without losing track of real-world operations.

From CapEx to OpEx: what is actually shifting?

Moving from CapEx to OpEx is not only an accounting preference—it changes how IoT is designed, procured, and governed.

IoT becomes a product, not a project

In a CapEx mindset, success is deployment completion. In an OpEx mindset, success is service performance over time: uptime, security posture, data quality, response time, and measurable business impact. This pushes teams to define SLAs, operational ownership, and continuous improvement from day one.

Procurement shifts to lifecycle contracts

More enterprises in 2026 want bundled offers (device + connectivity + platform + support) or managed services. This can simplify operations, but it also introduces vendor lock-in risk. A TCO model should explicitly quantify the trade-off between reduced internal OpEx and reduced flexibility.

FinOps meets IoT

Cloud governance practices—usage tagging, budgets, unit economics, anomaly detection—are increasingly applied to IoT workloads. If the platform bill cannot be explained in cost-per-device terms, it becomes vulnerable to budget cuts.

Key cost drivers that often get underestimated

1) Firmware and fleet update operations

Firmware management is rarely “free.” It requires staged rollouts, testing across variants, monitoring, and sometimes field remediation. The more heterogeneous the fleet, the higher the recurring engineering and QA effort.

2) Data egress and cross-region architectures

As deployments go global, data movement can become a structural cost. The cheapest architecture for ingestion is not always the cheapest for analytics, compliance, and long-term storage.

3) Truck rolls and physical reality

Battery replacement cycles, harsh environments, calibration needs, and site access constraints can dominate TCO in industrial contexts. TCO models should include realistic failure rates and service visit probabilities—not optimistic assumptions.

4) Security as an operating cost

Security expenses extend beyond initial device hardening. Certificate lifecycle, vulnerability scanning, security monitoring, penetration testing, compliance audits, and incident response are recurring, and they scale with fleet size.

How to build a defensible IoT TCO model in 2026

Step 1: Define the scope and time horizon

Most industrial IoT fleets should be modeled over 3–7 years. Shorter horizons hide recurring costs; longer horizons require explicit assumptions about refresh cycles and platform evolution.

Step 2: Create a unit economics baseline

Build a “device-year” baseline that includes:

• connectivity recurring cost
• platform consumption cost (average + peak)
• device management and OTA operations allocation
• security operations allocation
• support allocation

Step 3: Add scenario bands, not a single number

Use at least three scenarios:

• Expected: realistic adoption and operational maturity
• Constrained: higher failure rates, slower deployment, tighter compliance
• Optimized: improved automation, better device reliability, reduced cloud spend via edge filtering

Step 4: Separate “must-have” costs from “choice” costs

Some expenses are non-negotiable (secure provisioning, monitoring, incident readiness). Others depend on design choices (edge compute level, analytics depth, digital twin fidelity). Modeling them separately clarifies where architecture decisions matter most.

Step 5: Model scale effects and operational automation

At 100 devices, manual processes can work. At 50,000 devices, they become cost bombs. Add explicit assumptions on automation maturity—especially for onboarding, monitoring, and updates—and link them to headcount requirements and tooling costs.

Example structure: what your spreadsheet should look like

A usable TCO model does not need to be complex, but it must be structured:

• Inputs: fleet size growth, device types, average data volume, connectivity profile, SLA level, failure rate assumptions.
• Cost blocks: CapEx (one-time), OpEx recurring (monthly/annual), OpEx variable (per message/GB), risk reserves (percentage or fixed).
• Outputs: cost per device-year, cost per site-year, total annual run rate, marginal cost of adding 1,000 devices, sensitivity analysis.

Once these outputs exist, IoT decision-making becomes significantly clearer: teams can compare architectures, suppliers, and rollout plans using the same economic language.

Interpreting TCO: where ROI gets real

TCO is only half of the equation; the other half is value capture. Many programs struggle not because the technology fails, but because operational ownership is unclear. If the organization cannot convert data into decisions—maintenance work orders, process changes, inventory optimization—then even a “cheap” IoT stack becomes expensive.

The most resilient business cases therefore combine:

• measurable outcomes (downtime avoided, energy saved, compliance hours reduced)
• repeatable operations (standard onboarding, monitoring, support processes)
• predictable unit economics (transparent cost per device-year)

What best-in-class looks like in 2026

Enterprises that scale IoT successfully tend to share the same playbook:

• They design for operations: device management, security lifecycle, and support are built into the architecture.
• They measure unit economics: cost per device-year is tracked like any other operational KPI.
• They constrain data: edge filtering and event-based telemetry reduce cloud spend without losing insight.
• They industrialize change: firmware updates and configuration changes are treated as disciplined release processes.

Conclusion: the IoT winners will be the ones who master cost-to-serve

IoT is no longer judged by pilot success. It is judged by scalability, resilience, and the ability to operate safely over years. A modern TCO model is the bridge between engineering and finance: it makes costs explicit, reveals risk multipliers, and supports architecture choices with defensible unit economics.

As IoT shifts from CapEx-heavy rollouts to OpEx-driven services, the winners will be the teams that can answer a simple question with confidence: “What does it cost us to run this fleet for the next 12 months—and what will it cost when we double it?”

The post IoT Total Cost of Ownership (TCO) Models: From CapEx to OpEx in 2026 appeared first on IoT Business News.

• Katie Davis, now 21, started experiencing waves of intense stomach pain when she was 20.
• Doctors initially suspected a benign ovarian cyst and told her to monitor symptoms.
• After she got chills and nausea, she got a colonoscopy and was diagnosed with stage 2 colon cancer.

Katie Davis lived the typical busy college student life when she first started feeling stomach pain.

Then a 20-year-old junior and marketing major at Westchester University in Pennsylvania, Davis split her time between classes, her job at Playa Bowls, and her sorority. The pain in the top right of her abdomen was easy to ignore because it was so sporadic and fleeting.

“It was on and off, it would come in waves,” Davis, now 21, told Business Insider. “I would go a good while without it, and then it would come and only last a few minutes, sometimes even a few seconds.”

Over time, the pain — when it showed up — got more severe, sometimes causing her to double over in pain. Three months after it started, she went to a local urgent care while at her boyfriend’s family beach house. There was no ultrasound equipment at the facility, and she was told that, based on her symptoms, it could be an ovarian cyst that would hopefully go away after her next period.

Her doctor suspected colon cancer before the biopsy

Davis was told to keep an eye on the pain and go to an emergency room if she felt other symptoms like fever or nausea. A few days later, when she started getting chills and vomited at her parents’ home, Davis did just that.

“That was the first time anything more serious than an ovarian cyst was brought up to me,” Davis said. According to her ultrasound and CAT scan, her colon was inflamed and appeared to have free fluid, a potential indication of infection, trauma, or cancer.

The ER doctor thought it could be Crohn’s or ulcerative colitis, or, in more serious cases, colon cancer. He scheduled a colonoscopy.

“I didn’t think I was going to come out of it having cancer or anything really serious,” Davis said. Her first clue that something was wrong was when she woke up from the procedure. She noticed she was placed in a separate room from the other colonoscopy patients.

The doctor who performed the procedure told Davis and her mom that he was “pretty positive” the mass in Davis’ colon was cancerous. “He said he’d been doing it for long enough that he could kind of tell,” she said.

Shortly after, Davis was diagnosed with stage 2 colon cancer.

“I didn’t really know what to think or feel,” Davis said about learning her diagnosis. “Definitely just numb and confused at first, like ‘how did I get this?'”

Treatment dragged on due to side effects like vision loss

After diagnosis, Davis had surgery on her colon and was supposed to start three months of chemotherapy soon after. But, the side effects complicated her treatment.

“I couldn’t tolerate the more hardcore chemotherapy,” Davis said. She developed extreme fatigue, nausea, and neuropathy, which she said felt like “pins and needles” in her hands every time she encountered temperature changes.

The most alarming side effect was her vision loss. “My vision would go completely black,” Davis said. Her parents researched the drug, oxaliplatin, which can cause vision issues in some patients. Davis also found the Colorectal Cancer Alliance (CCA) and said hearing similar stories around common side effects helped her stay informed about alternative treatment options.

Davis was put on oral-only chemotherapy medication, prolonging her treatment from three to six months. The only upside was that she no longer had to travel back and forth for treatment, since she could take it wherever she was.

All the while, she was still attending her college classes in person as often as she could, even though her professors knew about her colon cancer diagnosis. “My boyfriend lives there, all my friends live there, so I tried to be there as much as possible,” she said. “I tried to keep up with my stuff as much as I could, but it definitely was difficult to do schoolwork when I felt as horrible as I did on the chemo.”

She’s glad she listened to her body

Davis finished chemo in June 2025 and was declared cancer-free shortly after. Going forward, she’ll have blood tests every three months and an annual colonoscopy.

Now a senior, she’s a marketing intern at a financial advisory firm and is trying to figure out her plans post-graduation. She said finishing treatment made her feel “excited to be normal again” and get back to her normal college life without worrying about doctor’s appointments or treatment side effects.

Looking back, she’s grateful for noticing the warning signs early enough. “A lot of my doctors said that most people at my age or with my stage wouldn’t really have the symptoms that I had that let me know that something is wrong,” she said. “I’m glad that I learned to listen to my body.”

It’s her biggest piece of advice to young people with similar or subtle symptoms, as colon cancer recently became the leading cause of cancer death in people under 50. She said joining the CCA and colon cancer Facebook groups can also help raise awareness of potential symptoms.

“You’re not really alone going through it,” she said, whether you’re worried about symptoms or actively undergoing treatment. “There are other people who are experiencing it too who can help you.”

• My son bought a $2 car with no title and major damage.
• He spent two summers fixing it himself using YouTube.
• The independence he gained was worth more than the car.

My eldest felt a strong urge to own a car for most of his teenage years. He would pop into the living room and show his dad and me his latest internet find, usually a 20-year-old jalopy with questionable reliability costing several thousand dollars.

Each summer break, he would talk about buying a car with us. Each time, we wouldn’t say no. We would just urge him to consider his situation as a full-time student with an uncertain future.

We provided transportation to and from school, and while there, he walked, rode a bike, and grabbed rides from friends. Each time, he decided on his own that it might not be the smartest time to invest a couple of thousand dollars in a car of questionable repair.

He got a deal from a family friend

The summer before his junior year of college, however, a family friend offered him a deal he couldn’t pass up. It was a 20-year-old Volvo wagon that had a run-in with a deer. The front end was crumpled, it was undrivable, and he didn’t have the title. But the price was right — $2, twice the friend’s original purchase price of a buck.

In some ways, it was a bold move. My son didn’t have much experience in car mechanics. During that school break, he put money and time into repairs. He straightened the front radiator support with a winch so all the parts would fit again. He replaced the radiator and flushed out the intercooler. By the end, it stayed in motion long enough to limp to a storage barn for the winter.

The next summer, it was his main focus. He installed a new radiator fan, bought a new battery. Replaced two tires and had them aligned. Put in a new headlight and did more bodywork. He cleaned it, inside and out. And just a couple of days before returning to college, the crowning glory: a salvaged hood that perfectly matched the golden hue of his car.

He learned a lot from fixing the car

There was a fair amount of angst. Figuring out the process for issuing a new title. Hunting down the owner who last had it and arranging a meeting. Ordering the wrong or incomplete parts and having to send them back. Determining what needed to be fixed and how much it cost. Calculating how much he should spend, even after fixing it up, the car was probably only worth about $2,000.

He elected to do much of the work himself, spending hours at the “University of YouTube.” At one point, as he lamented the money he had spent so far, with the possibility that it would all be for naught, my husband asked him how much a college credit hour costs. My son looked it up. It was exactly what he had spent so far on the car. My husband said, “Haven’t you learned a lot?”

That helpful reframing stuck: The night before he drove the car to college, my son commented, “Hey, I got a free car at the end of that college class.” We celebrated with him that evening, telling him how proud we were of his persistence and frugality, of his push to learn something new.

He got the independence he wanted

Seeing him drive off in that car left an indelible impression on me. Armed with his insurance’s roadside assistance, a toolbox gifted by his dad, and a bag of extra fluids in the passenger seat, he set out on the 8-hour, 57-minute drive to York College in Pennsylvania from our home in Kentucky. He couldn’t shake the small, satisfied smile on his face. I couldn’t shake my delight and my apprehension.

Being the mom that I am, I asked him to text whenever he stopped so we could track him on his journey. First stop: at the coffee shop halfway, our usual lunch break, and the new thrift store next door. Next, at a Civilian Conservation Corps museum, he saw signs along the highway. Finally, in the parking lot of his dorm. Even through text, I could sense the satisfaction and pride he felt for accomplishing that trip in his own ride.

In the ensuing months, the $2 car has safely delivered him each week to his internship and to a friend’s house for fall break. It has given him a measure of independence he didn’t have before. And it gave him something we, as parents, couldn’t, no matter how much we wanted to: a sense of self-sufficiency. That was something he had to earn.

We could only encourage him, support him, and talk him through his next steps, then see if he succeeded or failed. In the end, he knew that he could handle the road ahead by himself.