Author: admin

• Coinbase CEO Brian Armstrong talked about some of Coinbase’s strategy for keeping token costs “roughly flat” on X.
• “We’re working hard on routing prompts to cheaper models where appropriate,” Armstrong wrote.
• The post ignited debate over token-saving strategies — and what it means for the big AI labs.

Not every AI prompt needs Opus 4.8.

As the fervor of tokenmaxxing dies down, some AI users are wondering how to get more bang for their buck and keep their monthly costs in check. Coinbase CEO Brian Armstrong shared the crypto company’s strategy: not skimping on the cheaper models.

“We’re working hard on routing prompts to cheaper models where appropriate, and in some cases have been able to keep costs roughly flat, while token usage continues to grow exponentially,” Armstrong wrote on X on Sunday.

While the latest models like Opus 4.8 or GPT-5.5 promise bleeding-edge benefits, they can also devour more tokens. (That’s before you turn on Fast mode.) When Anthropic launchedi Opus 4.7, many users complained that they were quickly hitting rate limits.

Armstrong wrote that he anticipated “80% of workloads will be running on 99% cheaper models within 12-18 months.”

The only times when users will use the latest models, Armstrong predicted, are when they need to be “IQ maxing.” This includes scientific breakthroughs or agent orchestration.

“This leads me to think the limiting factor will be energy and compute, not better models,” Armstrong wrote.

The Coinbase CEO’s post caught the attention of some tech luminaries. Venture capitalist Marc Andreessen called it “interesting.” Hugging Face cofounder Julien Chaumond wrote that “model routing is growing a lot these days.”

Box CEO Aaron Levie wrote that Armstrong’s numbers were a “bit extreme,” but that AI use would likely stratify in the coming years. “High end” work will be completed by leading models, Levie wrote, while “high volume” work will be relegated to the cheap models.

“Intelligence allocation is going to be extremely important,” Harvey cofounder Winston Weinberg wrote.

The efficiency mindset is relatively new — or at least new to publicly flaunt. Not long ago, when tokenmaxxing was all the rage, tech leaders would post their high token bills or flex their usage of the latest models.

That mindset is especially popular in the startup space, where Y Combinator CEO Garry Tan advises founders to “let it rip” with tokens. Lance Yan, a YC-backed startup founder, told Business Insider in April that rationing tokens was “stupid.”

The tide seems to be turning. Glean cofounder Tony Gentilcore commented that Armstrong’s post was “spot on.”

“Everyone technical already knows this,” Gentilcore wrote. “The financial markets are the only ones extrapolating out Opus prices to infinite scale.”

• Founders are publicly airing their “nightmare” interactions with venture capitalists.
• They shared on X what they say have been some of their worst experiences with investors.
• Legendary VC Vinod Khosla joined the fray.

Over the past few days, some Silicon Valley tech founders have taken to the internet to publicly roast the venture capitalists they once pitched.

The kefuffle began last week when Greg Isenberg, the host of “The Startup Ideas Podcast,” posted to X about his experience raising a $15 million Series A.

“12 people in the meeting. One of the GPs fully fell asleep. Out cold for 30+ minutes. Nobody acknowledged it. Everyone just kept going,” Isenberg wrote, referring to an unnamed general partner.

Isenberg said he continued his presentation, sharing slides with an investor he called an “unconscious man in a Herman Miller chair.”

“That’s venture capital,” he wrote.

Venture capital is the financial engine of the tech industry. Founders with an idea rely on them to kick-start or turbocharge their fledgling companies. Typically, this is not a loan, but an equity exchange. The earlier a VC invests, the riskier the investment. When a company succeeds, however, the returns can be astronomical.

Uber founder Travis Kalanick, responding to Isenberg, said that the venture world has changed over the years. Investors, he said, once took a more casual approach to pitch meetings.

In 2001, Kalanick said he pitched an investor who was sitting in his “parked Lexus.” Kalanick was sitting in the passenger seat.

Kalanick said the investor “grabbed” his laptop, placed it “on his large belly,” pressed it against the steering wheel, and began flipping through the slides himself.

“2001 fundraising hit different,” he said.

The back-and-forth went viral among the niche community of already-successful millionaire and billionaire founders who are terminally online, and the pile-on began.

Cloudflare CEO Matthew Prince said a Sequoia partner passed on Cloudflare because “he didn’t think a woman could lead a security infrastructure company.” Cloudflare was founded in 2009 by Prince, Lee Holloway, and Michelle Zatlyn. The internet services company is now valued at almost $90 billion.

Prince said he also once met with Khosla Ventures about investing in Cloudflare’s Series C.

Vinod Khosla, the legendary tech investor and the firm’s namesake, took Prince and his fellow founders out to dinner, Prince wrote. Near the end of the conversation, Prince recalled, Khosla leaned over and said, “I’m impressed with you, not so much with them, what if you fire them and I’ll give you all their stock?”

Prince said he was so offended he never spoke to Khosla again. Other founders responded to Prince, adding their own experiences dealing with Khosla.

It was enough for Khosla to spend his Saturday responding to the accusations, firing off over a dozen X posts. In some, he denied the stories and asked for evidence, but in most, he repeated a single mantra: Honesty is the best policy.

“I am often wrong but always give honest opinions. Some find this harsh, but hypocritical politeness hurts founders,” he wrote in one post. “Brutal honesty gives them a chance to evaluate it and accept or reject the opinion. Great founders elect for honesty. It is not fun to offer brutal honesty.”

Khosla did not immediately respond to a request for comment from Business Insider.

Some industry insiders came to Khosla’s defense. Blake Byers, an early-stage investor and founder, said Khosla founded one of the pioneering companies of the modern computer industry — Sun Microsystems — before becoming one of Silicon Valley’s most influential venture capitalists.

“He is one of the truest VCs to ever do it,” he wrote.

Mark Cummins, an angel investor and robotics expert, said he began pitching a partner at a French firm, only for the investor to interrupt with questions about his parents’ careers.

“What did your father do?’ the partner asks me in a thick French accent,” Cummins wrote.

When Cummins said his father had trained as a theoretical physicist before going into business, the partner replied: “Aha! Your father was a failure!” Then, after Cummins said his mother had been a biochemist before becoming a schoolteacher, the investor said, “Also a failure!”

Cummins said he responded, “I have a hundred employees and we need funding. ‘Would you like to hear about my company?'”

In a conversation largely dominated by men, Claire Vo, the founder of ChatPRD, an AI product management platform, said an investor once interrupted her pitch to say he was glad she wasn’t trying to have kids while building a company.

“I love an opportunity to tell a nightmare VC story!” she wrote.

She later turned one of Khosla’s posts defending himself into what she called a “pop punk banger.”

• I started teaching my daughter to pack when she was 3 years old.
• Packing together helps build her independence and decision-making.
• I hope the extra effort now will pay off as she grows older.

I started teaching my daughter to pack when she was 3 years old, which, I know, sounds inefficient at best and slightly unhinged at worst.

Packing with a preschooler is slow. It turns a 30-minute task into a 90-minute one. It involves delicate negotiations over which stuffed animals are “essential,” last-minute outfit swaps as she re-discovers her favorite sparkly boots, and frequent distractions. If saving time today were the goal, I would simply do it myself.

But my perspective changed after I heard versions of the same complaint from multiple mothers; they were still packing for their teenagers. Not just occasionally helping or reminding, but fully responsible for it. That dynamic doesn’t appear suddenly at 13; it builds over time.

So I decided to start while my toddler still craves independence with a fervor, hoping it will pay off over the next decade.

The first step was participation, not decision-making

The first time we “packed together,” I did almost all of the work in advance. I pulled everything she would need and laid it out on the floor of her room, alongside an open suitcase. My toddler wasn’t choosing items or deciding quantities. She happily folded clothes with me, shoved them into packing cubes, and put the packing cubes into the suitcase.

This participation also helped her mentally prepare for the coming travel. As we packed bathing suits, we talked about going to the pool and the beach. Her blankie gets packed because we’ll be sleeping in a new place, so she’ll want to have something familiar. She’ll wear her sneakers to the airport because we’ll be walking more than usual.

Especially with toddlers, a smooth trip starts with helping them understand what the experience will be like before we get there. That mental preview reduces friction later.

The next step was constrained choice

Once that baseline was established, I shifted one variable: selection. Instead of laying everything out myself, I told her what we needed, and she got to pick: six T-shirts, five pairs of shorts, two bathing suits. Then she went to her drawers and chose them.

That changed the task meaningfully. She still wasn’t determining quantities or planning for contingencies, but she was making decisions within a defined framework. I separated “what do we need?” from “which specific items do we bring?” and introduced them in sequence.

She’s also responsible for helping put everything into her suitcase, which makes it easier to say no to the constant requests to bring additional toys and books. For shorter trips, if it fits into her carry-on after her essentials, she can bring it. On longer trips with a checked bag, she gets one packing cube for toys and books so she can decide what she most wants to bring.

One major step is that now she has a suitcase that is clearly hers. For our coming trip to Asia, she’s packing into a light pink MiaMily suitcase that I gave her for her most recent birthday, now enthusiastically decorated with stickers. It’s a ride-on suitcase that she sits on proudly, like her travel throne.

That shift sounds small, but it changes how she approaches the task. It’s no longer a shared family suitcase or something I’m managing on her behalf—it’s her suitcase, and she treats it that way.

The goal is a gradual handoff

I’m not trying to create unrealistic independence at 3. Over time, I can shift more of that responsibility to her by asking her to suggest quantities, to think through activities, and to identify what might be missing.

It does take longer now. There’s no way around that. But if the alternative is still packing for her a decade from now, the tradeoff is worth it.

• My dad planned to travel and do so much once he retired, but he died at 52 before he had the chance.
• Losing him pushed me to live life differently, like skipping college to travel to Europe instead.
• I’m grateful I live a life focused on new experiences and adventure, not stability and savings.

Throughout his life, my dad was incredibly focused on being successful and providing for our family, trying to ensure all of us had the best possible future.

He was good at it, too. He built us a beautiful home by the beach, studied with me so I’d get straight A’s, and took our family on countless vacations to Hawaii and Yosemite — all while running a business and putting money into his retirement fund.

Along the way, he postponed a lot of things he wanted to do, like finally taking that trip to Europe, buying his dream car, and relocating to Hawaii with my mom so they could spend their days sipping mai tais under the shade of a palm tree.

It was OK, he’d say, because once he retired, maybe in his 60s or 70s, he’d do all of those things. But then he got sick at 51. One year later, at 52, he was gone.

I was only 15 then, and I didn’t know how to handle losing the bravest man I knew. It was so unfair, and life suddenly felt incredibly short.

My dad planned to travel and experience so many grand things after retirement, but never got the chance. I didn’t want the same to happen to me.

His death changed my priorities and how I thought about the future

Losing my dad left me yearning to experience things he didn’t get to. I didn’t want to wait until an uncertain retirement; I wanted to live my life now.

After high school, I didn’t end up going to college, which, before my dad’s death, was the only option I saw for my future. Instead, I started working and saving up money to travel. By the following summer, I was able to make my way to Europe.

As I ate fresh slices of pizza in Naples and took in the sights from the top of the Eiffel Tower, I felt like my dad was right there with me.

That trip gave me a taste of a life I wanted to live.

Shortly after, I decided to pursue teaching English abroad in Mexico. It was another chance to see more of the world, but starting from scratch in a new country and learning a new language wasn’t easy.

It took time to find friends, and I didn’t always have a steady paycheck — but what I didn’t make up for in my savings account, I was making up for in life experiences.

After a few years, I had become fluent in Spanish, traveled throughout Mexico, and found a passion for writing and playing guitar. My confidence grew, and I felt I could do this somewhere else.

So in my early 30s, where years prior I would’ve thought I would be married with children, I packed up and moved to Europe, single, with some savings and no steady job waiting for me on the other side.

I landed in Spain and started building a life from scratch again in Barcelona. It was incredibly lonely at times, and I often compared myself to friends back home in Los Angeles, who had steady careers and were beginning to have kids.

But then I would think of my dad and how he’d always tell me I could achieve anything I put my mind to. I reminded myself that I chose this path. I was seeing the world and making my way in it, even if it wasn’t always easy.

I’m glad I’ve chosen to plan a life around experiences, not stability

I’m nearly 40, and I’ve been in Barcelona for almost seven years now. In that time, I’ve built a community and met and married the love of my life. I don’t have children yet, and my writing career is still blossoming.

I still haven’t prioritized growing my savings account or making concrete retirement plans. It can be a risky way to live, and sometimes I’m scared of what the future may bring, but losing my dad so young made me realize that nothing in life is guaranteed.

I’d rather enjoy the present and grab opportunities as they arise rather than wait and potentially never get to take them — and I’m grateful for everything my dad taught me, even if I only had him for 15 years.

• I got rare access inside Intel’s factory, where it makes some of the world’s most advanced chips.
• The plant is so tightly controlled that paper, white light, and even footsteps can become problems.
• The trip made me realize that making the chips powering everyday life means protecting them from us.

I knew visiting Intel’s chip factory would be different when they told me I couldn’t wear my regular deodorant.

Or lotion. Or hairspray. Or makeup.

Before I’d even boarded the plane to Oregon, Intel sent my videographer and me a long list of things we couldn’t wear or bring into its factory in Hillsboro. No Velcro. No Bluetooth. No phones unless they were on airplane mode. The list kept going.

That was my first clue I was about to step into a place governed by a very different set of rules.

In March, after months of planning with Intel, I got rare access to one of its chip factories — the kind of place the tech industry calls a fabrication plant, or fab. Inside, Intel makes some of the most advanced semiconductors in the world.

Chips run almost every part of modern life: laptops, phones, chatbots, washing machines, fighter jets, and the data centers behind AI.

Demand for these chips is skyrocketing, with annual semiconductor sales expected to reach $1 trillion by 2027. I went behind the scenes to see the complicated and delicate manufacturing process that’s so controlled, it permanently changed how I think about what it means to be clean.

What it takes to get inside Intel’s fab

I arrived on a rainy Oregon morning in my best walking shoes, per Intel’s instructions. We’d be covering a lot of ground, and they weren’t kidding. The fab in front of us was enormous — bigger than an aircraft carrier.

We walked about 10 minutes to the gear-cleaning room. Just beyond it sat a room full of what I estimated to be billions of dollars in Intel chips.

“Each little tiny speck can cause a defect, which would destroy the chip,” Chris Auth, Intel’s vice president of manufacturing development and our guide for the day, told me.

We scrubbed down every piece of camera equipment with sterilizing wipes. Not just the obvious surfaces. We extended the tripod legs, wiped them down one by one, collapsed them, and wiped them down again, hunting for any nook that might be hiding dust.

Then came the gowning room, a chamber so big it could have swallowed my New York studio apartment many times over. It was packed wall-to-wall with bunny suits, each worth about $1,000, according to Intel.

I got to wear one for the day. But first, I had to figure out how to put it on.

“So you kinda wanna scrunch up your suit so that the sleeves don’t touch the ground here,” Auth said, once he’d slipped on his hood.

Every piece had to connect in the right order. The onesie snapped onto the hood. The boots attached to the suit. My first pair of gloves got tucked under the sleeves. A second pair went on top to trap the skin particles my hands were shedding.

I’ve visited factories before that worried about visible contaminants — a bracelet falling off, an earring coming loose. Intel worried about invisible ones. The kind your body releases constantly without asking permission.

That also explained why my notebook had to stay outside. Regular paper sheds microscopic particles, and here, even that can be enough to ruin a chip. Intel handed me a special cleanroom notebook that doesn’t shed.

Then I stepped onto the fab floor.

Intel’s most precious room

In a place this tightly controlled, I was oddly thrilled by what looked like hot-pink equipment everywhere. But it wasn’t actually pink.

The gigantic room glowed under yellow light to protect the chips. Any other wavelength could damage a chip while it’s being made.

“Under yellow light, everything that looks pink to your eyes is actually red,” Tyler Osborn, Intel’s director of advanced packaging technology development, later told me, gently bringing me back to reality.

Nothing about the fab felt quite real, though.

There were more robots than people. The few people who were there all looked the same in hooded suits. Employees told me they often recognize one another by how they move.

“You get to know people’s gait,” Osborn said, his voice muffled by the layers covering everything but his eyes.

Robots zipped by on overhead tracks, carrying sealed boxes of wafers — the thin slices of silicon that chips are built on — around the factory, keeping them out of human hands.

People, I learned, are too inefficient for this work. Robots need to move thousands of wafers a day. Not to mention, humans can be clumsy.

I couldn’t stop wondering what would happen if someone in a rush tripped and sent a box of wafers flying.

“Mistakes are very, very costly,” Auth said. “You’re somewhere in the $50,000 to $500,000 range just for one wafer.”

Each robot carries 25 wafers at a time.

“So now you’re into the millions for just one box,” he said.

Even my footsteps felt like a risk

The chips are about the size of a fingernail. But Intel isn’t just controlling that tiny patch of space. It is trying to steady an entire factory around something microscopic.

“We’re building the world’s smallest features in some of the world’s biggest factories,” Auth said.

The fab is built in layers, with a foundation designed to absorb outside shocks — earthquakes, nearby machinery, even low-frequency vibrations from air-conditioning units in neighboring buildings.

“It comes down to microvibrations,” Bob McMillan, Intel’s life safety and systems manager, told me.

That was the moment I became unusually aware of my own footsteps. I felt like a giant moving through a world built for things far smaller and more delicate than me.

Everything in the fab was choreographed to protect the chip, which made me wonder what would happen if all that control failed in the smallest possible way.

So I asked what would happen if a beard hair got into one of these machines.

“You’re done,” McMillan said.

“A hair is huge,” Auth later explained.

A single human hair can be a million atoms thick. Some of the structures Intel is building are just a few atoms wide.

The entire factory works like a machine

It was hard to hear anyone over the constant hum of gigantic tools.

Then I realized the building wasn’t just full of machines. It was one.

Even the floor was working.

It stretched beneath us like a giant metal sieve, perforated with holes as far as I could see. They were there to pull particles away from the wafers — catching anything rogue that escaped our suits in less than 60 seconds.

“We change all the air in this factory about that quickly,” McMillan told me.

They filter the air over and over because, at any given moment, there can’t be more than eight particles bigger than a micron floating in every cubic meter of air. The room you’re sitting in right now probably has millions.

To me, that was hard to fathom.

I was standing inside one of the cleanest places on Earth, which was reassuring and also vaguely unsettling. Regular life suddenly seemed impossibly filthy.

Why making chips is so hard

A single chip takes about three months to make, and almost nothing can go wrong. In that time, it moves through roughly 2,000 steps, and thousands of machines spread across the factory.

“There’s 12 football fields of clean room space here,” Auth said.

He later told me it costs about $20 billion to build a fab like this. By comparison, One World Trade Center cost about $3.9 billion to build.

Despite the price tag, the US government has made building chips in America a top priority.

About 90% of the world’s most advanced chips are made in Taiwan, which Washington sees as a major geopolitical risk as China threatens to take the island by force.

That’s why, no matter how hard these chips are to build, the White House says the US needs more factories like this one. Right now, Intel is the only American company designing and manufacturing advanced logic chips on US soil.

I left thinking about how fragile modern life is

When we got back to the gowning room and took off our hoods, I realized I’d almost forgotten what everyone we’d spent the day with actually looked like.

Then I got my phone back, stepped back into normal life, and had a thought I still can’t shake: we live in a world that runs on chips.

To make them, however, we have to create an entire environment designed to protect them from us.

• Cara Katz is a 36-year-old working mom to a toddler.
• She has been using AI at work for at least six years.
• She said people shouldn’t judge moms who use AI — they aren’t the ones to blame for the ethical concerns around it.

This as-told-to essay is based on a conversation with Cara Katz. It has been edited for length and clarity.

A little over a year ago, AI became more prevalent in my mom communities than it had been in years prior.

I had been familiar with and using AI for about six years at work, where there was a lot of skepticism about its use — and back then, it wasn’t something we used at home. Fast-forward to about 18 months ago, and parents started using it to increase efficiency, saving time and maximizing output in their personal lives.

At first, it was mainly parents in tech and marketing, but now it’s being used by parents with no tech experience. We didn’t feel any shame about it at first, but things have changed thanks to the explosion of public opinion about using AI.

Since I’ve started using it at home, I’ve saved myself at least 10 hours a week. I’m tired of being judged for it.

I use it to plan our week

Scheduling is my favorite way to use AI at home because it saves me the most time, though it can be slightly complicated.

I used to spend most Sundays planning our week as a family of three. I work from home, and my husband works from home two days a week. My daughter goes to transition school, which is two hours of school two days a week. We don’t have any organized childcare, but she does have enrichment activities like music and gym.

My husband and I take turns caring for our daughter, but if both of us are busy, I plan for a babysitter to have her in the house or to take her out.

It makes for a really busy schedule to arrange and remember.

I started using Claude Code, which sounds scary, but it’s just a chatbot for code writing. It can be used exactly the same way as Claude.

It walked me through how to organize our Google calendars — our daughter’s calendar and our separate work calendars. Claude reads them all, and I prompted Claude to create this beautiful, color-coded HTML calendar.

I also fed it my daughter’s periphery schedules — like library events for the month — and trained it on her preferences. Every week, it sources events that she would be interested in, and puts them into the calendar.

It sends our babysitters’ recommended times, based on their previous work-time preferences, and asks them to agree to the date and time provided to have our daughter.

We then publish the schedule on Netlify, press “deploy,” and it creates a password-protected website that caregivers can view. If there is any chance, I enter that change, and it automatically updates and emails everyone a link leading us back to the website.

These days, I spend five minutes here and there on scheduling.

It also plans my shopping list

AI does all of my grocery planning. It runs a full inventory of my pantry to know what we already have, knows all of our preferences, knows my daughter and husband are celiac, and even knows my husband’s blood results. It uses all of this information to design a shopping list and meal plan.

I connect Claude to DoorDash and Uber Eats to get our groceries delivered. You can set it up so that this happens automatically, but I like to have a look at the list before I pay.

I followed my daughter’s developmental milestones with AI

I think moms get freaked out by developmental milestones — we know milestones exist, but aren’t taught that there is a range within these milestones.

When my daughter was a baby, I built a Claude project that researched which milestones she should be hitting. I asked for an updated list of activities we could do each week to help her achieve her developmental milestones. We printed it out and ticked off the activities.

When we went to the pediatrician for a check-up, we knew where she was developmentally before we even walked through the door. We could then have empowered conversations with the doctor.

I get more time with my kid

I love going back to my mom communities after being pelted with strong opinions from people who say that hate AI. We are buying back time with our kids using AI. My mental workload has lightened.

It’s hard for me to take people who are angry about AI seriously. It is easy to be theoretically morally angry.

It is quite different to sit across from a single mom, stay-at-home mom, or a working mom who struggles to pay bills, or spend time with their family, and tell her not to use AI because it probably uses less water than big organizations do.