Leveraging the latest version of MLX, Muse Spark Open was optimized to run on Mac Studio with M5 Ultra, and together they’re the best way to run a powerful model on your own desktop.

Apple’s unified memory architecture lets the M5 Ultra's integrated GPU use up to 512 GB per Mac Studio, so two machines can hold an entire model. And the chip’s 1.2 TB/s of memory bandwidth is fast enough to read the small slice each token needs. The result is two Mac Studios behave like one supercomputer with a terabyte of unified memory.

We tested a range of agentic workflows running entirely on-device, using Muse Code powered by Muse Spark Open, including shipping real iOS changes to Ice Cubes (an open-source Mastodon client) — bug fixes, new features, full build — and driving DaVinci Resolve end to end, with the model reviewing footage frame by frame to find the strongest moments and crop them vertically.

Find transcripts of your
favourite podcasts with exact timestamps

Some examples of people quickly accomplishing ambitious things together.

I frequently proclaim (together with other people working on AI) that we ought to be more ambitious when using it. It’s become a rallying cry. Whenever we see something impressive, we repost and write “we’re not ambitious enough”. I understand how tricky it is to follow that advice, so here’s a tutorial on being more ambitious.

I think of ambition in four types: Speed, Altitude, Depth, and Range.

• Speed: The same work in less time, the bread and butter for AI’s economic value: If you need some code written, I’ll do it faster with Claude. Well established and requires little creativity in how you approach AI, so I’ll skip this one.
• Altitude: Going up in abstraction and considering the outcome rather than the individual task.
• Depth: Considering a task the way a field-leading expert would. Rather than operating at the task’s layer, we go one level deeper and touch the underlying data, tools, and problems.
• Range: Drawing on multiple fields rather than just the one the task might need at the surface, using knowledge, tools, and techniques you normally wouldn’t.

Over my 20-year career as a founder, executive, operator, advisor, and investor in enterprise SaaS companies, I’ve been through a lot of ups and downs.

I’ve been an executive at two companies at inception (founding employee then CEO at one, and founder/CEO at BetterCloud). I was also an executive at one of the first cloud consulting companies just as they were getting started. At these companies we generated $100s of millions in ARR and hired thousands of people, including dozens of executives. We raised rounds from the best VCs in the world. But on the flipside, I’ve also been through a number of macro downturns, company pivots, existential moments of impending company failure, cash crunches, and terrible hires.

Recently, I’ve taken a step back from a day-to-day operating role. I’ve been spending more of my time advising early stage founders, many of whom are facing those very same challenges and situations for the first time.

To help shorten that learning curve, I’ve put together this “operating manual” with 50+ of my biggest lessons learned over the past two decades. It’s filled with personal anecdotes, actionable tips, and advice to help other founders avoid the same mistakes I made. I plan to apply all these learnings at my next company. Feel free to pick and choose whichever ones are most relevant for yours. (Note: These lessons are aimed toward early and growth stage companies.)

Not everyone will agree with my learnings or suggestions. As a member of several founder/CEO groups, I know people have strong opinions on how to run a company. That said, I’m still hoping this will resonate with readers. If each person just takes away one insight that will help them grow their business, then this was a success.

If there’s one general piece of advice that underpins everything, it’s:

Building is hard. Stay in the game to give yourself a chance to win.

The average time to an exit, if there ever is one, is over a decade. Be prepared for the grind. Yes, it’s going to suck at times, but it will get better.

Three critical things will help get you through the long haul: resilience, perseverance, and conviction. Ride it out. Continually learn, evolve, and move forward.

Keep going, keep growing.

“In general, I’ll say there was not one POC in which we entered and didn’t find malicious stuff,” she said. “Whether those are extension packages or skills — some of them had active live attacks while we did the POC. Which again, for me, product-wise is amazing. For the companies, it’s not very good.”

Pluto’s entire bet is that securing such environments will require security tools to pay attention to what AI agents are doing in real time, in addition to keeping track of what and how employees are building and deploying.

“We’re moving from user behavior to agent behavior,” Bahat said.

In line with that vision, the startup’s software monitors the AI tools employees use, the skills, extensions, MCP servers, and other components those tools rely on, as well as the apps and agents being created. The company says it can identify risky components, spot potentially dangerous agent activity, and enforce security policies without stopping employees from using or making AI tools.

To do all that, Pluto connects to endpoint security and device management systems and can deploy its own sensors through them or directly onto devices. It gathers information, including prompts, commands, running processes, permissions, and users’ previous activity, and uses its own “lightweight” models on endpoints to help assess what an agent is trying to do.

What stood out is that Pluto’s system can intervene while an agent is working, Bahat said. If an agent tries to gain permissions beyond what a user intended, for example, the system can stop the action or add instructions into the session that prompt the agent to ask for confirmation.

“We can inject context back to the session, and this way I can change how the agent is working,” Bahat explained.

The startup also takes an unusual approach to finding customers. In April, it launched Plutonium, a free platform that lets anyone check risks associated with publicly available AI skills, connectors, and add-ons. There’s also an MCP scanner and a catalog of verified components. Bahat claims thousands of security professionals use Plutonium and that the community around it has become an important source of feedback and prospective customers.

US Vice President JD Vance said in March, without offering evidence, that the administration believed Omar “definitely committed immigration fraud”. Her office called the claim fabricated and defamatory.

Last month, Homeland Security Secretary Markwayne Mullin told conservative podcaster Benny Johnson that “something’s going to happen” in the case.

In early December last year, Trump took aim at the Somali community in Minnesota and called Somali immigrants “garbage“. “We’re going to go the wrong way if we keep taking in garbage into our country. Ilhan Omar is garbage. She’s garbage. Her friends are garbage.”

In January, a man sprayed her with vinegar at a town hall in Minneapolis. Anthony James Kazmierczak, who admitted attacking her because he disagreed with her progressive views, was sentenced last month to 14 months in prison.

Trump, meanwhile, returned to Omar’s origins at a rally in Youngstown, Ohio, on Monday this week.

“She comes from one of the worst countries anywhere on the planet,” he said, as the crowd chanted: “Deport.”

Moats are more important than ever in the AI era. Anyone can create a product; they can do it faster than ever before, and everyone is building the same things.

perhaps because it seems harder to measure, people often skip one of the most important criteria: Does it have a moat?

A company can check all your boxes and look like a rocket ship. If it doesn’t have a moat, though, it’s going to stall out. It won’t be able to keep growing once competition gets fierce

Moats are more important than ever in the AI era. Anyone can create a product; they can do it faster than ever before, and everyone is building the same things. But only some of these products will win.

The framework has three steps:

1. Identify a compounding advantage

2. Gauge how strong that advantage could be

3. Measure if it’s working

In reality, only a small subset of what people call moats actually drive the vast majority of long-term value.

They’re what I’ll call compounding advantages: not just a temporary edge, but a self-reinforcing flywheel that gets stronger and stronger over time, ultimately making a company very hard to compete with.

How to identify the compounding advantage

Your job is to identify which kind of compounding advantage a startup has the potential to build. It won’t be in full force yet (it takes a long time to develop), but it should be pretty clear which path they are on. That’s because it’s not something you can “bolt on”; it’s not some clever competitive tactic or a tweak you can make to your business model.

A compounding advantage is the core loop at the center of the business through which you add more value for customers. Defensibility emerges as a byproduct of turning the flywheel over and over.

So as you evaluate a company, don’t focus on what they say they are or what their “strategy” is. Focus on what they’re actually doing. What is the team working on every day? What’s on their roadmap? Which teams are large and influential inside the company?

Mike Vernal recently wrote something of a manifesto for anyone building this kind of business today:

If you are building a venture-backed software company, I think the answer is once again “do it all.” Just build the whole thing—your moat is the totality of what you’ve built and completely owning the buying center.

The bigger cost is what I’ll call the “workflow tax”: all of the time the team spends learning a product and building it into how they work, and how disruptive it is to do that over again.

At the core of these companies is a machine to both deepen and broaden customer engagement. The roadmap is rich with efforts to better serve existing use cases or create ways to serve new ones. They’ll often be focused on integrations and working with other parts of their customers’ stack. The team should be heavily tilted toward product and engineering, with an emphasis on shipping rapidly.

Network-effect products are common in industries like social (LinkedIn), gaming (Roblox), and marketplaces (Amazon).

The emphasis of a network business is less on building out the surface area or feature set of the product, and more on growing the network and making it easier and easier for that network to interact.

3. Scale

Businesses with scale as their compounding advantage are all about making it possible to offer the same product at a lower cost or a higher-quality product at the same cost.

4. Data effects

These are businesses with data at the core. It’s the input that makes the product higher-quality, more efficient, or work for more use cases. And by using the product, customers generate even more data, which further compounds this effect.

In these companies, you’ll often see a large effort to collect and sanitize data, and to build algorithms and products on top of it. ML and data science are usually large and influential teams.

AI Coding Agents: Turning Software Tasks Into Tested Changes

Coding agents turn software requests into a loop of investigation, editing and checking. A model chooses what to inspect or change; tools let it work in the repository; execution results shape the next step. Useful autonomy depends on understanding the task, preserving existing work and checking the exact change that will be integrated. This chapter follows that workflow without treating an agent’s completion message as proof that the software works.

AIE CODE returns for its second annual code-focused summit. Catch up on the year with the world’s top agentic engineers, sharing the workflows and breakthroughs changing how software gets built.

A year of progress moves fast. Get the full picture from the people driving it: the workflows they use, the systems they ship, and the lessons that changed how they build. Leave with ideas you can put to work.A year of progress moves fast. Get the full picture from the people driving it: the workflows they use, the systems they ship, and the lessons that changed how they build. Leave with ideas you can put to work.A year of progress moves fast. Get the full picture from the people driving it: the workflows they use, the systems they ship, and the lessons that changed how they build. Leave with ideas you can put to work.

This year, we’re also going deep on internal agents: how companies like Ramp and Coinbase build AI systems for their own teams. Alongside coding agents, expect code review, evaluations, and the infrastructure behind production workflows.

ntfy (pronounced notify) is a simple HTTP-based pub-sub notification service. It allows you to send notifications to your phone or desktop via scripts from any computer, and/or using a REST API. It's infinitely flexible, and 100% free software.

First-party residential, datacenter & mobile proxies. No expiry, no subscription, 24/7 human support

The hourly spot price for an Nvidia B200 GPU has more than doubled since March, according to data from Ornn, a startup that maintains indexes.

With Escrow.com you can buy and sell anything safely without the risk of chargebacks. Truly secure payments.

Domain Names

Escrow.com is the dominant payment method for the buying & selling of domain names, with transactions including uber.com, snapchat.com, spacex.com, twitter.com, instagram.com, freelancer.com, gmail.com, slack.com, wechat.com, chrome.com and wordpress.com. Learn More Motor Vehicles

When buying classic cars, a used sailboat or even an aircraft engine Escrow.com ensures money transfer and vehicle delivery with every sale. Our experienced personnel can even help you with shipping documentation, titles, liens and more. Learn More Electronics

Escrow.com handles the buying and selling large scale computer setups, professional sound systems and all manner of electronic equipment both big and small. Learn More General Merchandise

From computer hardware to luxury goods, you can safely and easily buy and sell merchandise all over the world with the protection of Escrow.com. Learn More Milestone Transactions

Paying for a good or service and want money released only at certain stages? Use Escrow to assure that money is released only when you're happy with each step. Learn More Jewelry, Watches, and Fashion

Buying and selling expensive jewelry online can be difficult, as it is extremely difficult to spot a scam. Escrow.com's simple 5-step process ensures money transfer and jewelry delivery with every sale. Learn More

Meeting
Transcription
API