weekend ai reads for 2026-03-06

Today’s startup competitors might move faster than their predecessors — but they’re much less prepared for the long term.

The opposition appeared overwhelming: Tens of thousands of emails poured into Southern California's top air pollution authority as its board weighed a June proposal to phase out gas-powered appliances. But in reality, many of the messages that may have swayed the powerful regulatory agency to scrap the plan were generated by a platform that is powered by artificial intelligence.

The strategy shift shows how Salesforce is still trying to figure out how to incorporate fast-emerging AI into its products. Benioff spent much of last year characterizing AI models from OpenAI and others as commodities while saying Salesforce had a pole position to sell AI to large companies because many already stored critical data on its servers.

At last year’s Dreamforce, he even took a shot at OpenAI, saying many customers were having trouble getting OpenAI models sold by Microsoft to produce accurate results, compared to Salesforce’s AI tools. (Ironically, Agentforce also relies in part on OpenAI technology).

Since then, more corporate workers have been using ChatGPT to increase their productivity. OpenAI has projected $400 million in revenue this year and $1 billion next year from corporate customers that use an enterprise version of the chatbot, though millions of other workers pay for individual ChatGPT subscriptions. Salesforce doesn’t break out Agentforce revenue, and Salesforce shares are down 29% this year on concerns about its ability to generate revenue and profits from AI.

While AI models can process large amounts of data, Khosrowshahi said that they still struggle to process and make choices based on new information — something that executives like him have to do.

Nobody gets promoted for the complexity they avoided.

The presence of a parent remakes a child, transforms them back into a child.

If there’s a takeaway beyond the spectacle, it’s this: the bottleneck in AI-assisted development isn’t the quality of your ideas - it’s the quality of your feedback loops. The games got dramatically better not when I improved the prompt, but when I gave Claude the ability to screenshot its own work, play-test its own levels, and lint its own scene files.

Authentic engagement is becoming increasingly important as AI increasingly commoditizes products and distribution. To go to market, we believe that a company’s only true competitive moat is brand belonging, community, and creating an authentic place for people to show up and be creative.‍

These modern systems are developed to sound human, natural, and conversational. Unfortunately this seems to come at the expense of accuracy. In my testing, both models had a tendency to skip words, read numbers incorrectly, chop off short utterances, and ignore prosody hints from text punctuation

You’ve spent months teaching another AI how you work. That context shouldn’t disappear because you want to try something new. Claude can import what matters, so your first conversation feels like your hundredth.

For years, you’ve sat in front of a rectangle, moving tinier rectangles, only to learn that AI can now move those rectangles 10x better. As someone outside the equity class, you begin to wonder what your role is in this new paradigm. And whether rectangles were ever your ticket to happiness in the first place.

“AI” has become semantically meaningless. The term now encompasses everything from a regression model to an autonomous robot, creating confusion in strategic discussions, partner conversations, and product positioning. This taxonomy provides a functional framework based on what the AI actually does, not what technique it uses.

Tensor is a specialised container for those numbers. You might know tensors from maths or physics, but in machine learning, a tensor is simply PyTorch’s data type for storing numbers. Think of it like a more powerful version of a list or array. Tensors hold your training data and the weights your model learns.

Newly released benchmark data illustrates just how aggressively these compact models are competing with—and often exceeding—much larger industry standards. The Qwen3.5-9B and Qwen3.5-4B variants demonstrate a cross-generational leap in efficiency, particularly in multimodal and reasoning tasks.

The defining characteristic of any CTrO—including an AI-first one—is a willingness to work toward their own obsolescence. Rather than accumulating scope or authority, they focus on embedding AI capabilities into the business, strengthening collaboration between technology and line leaders, and leaving behind an operating model that can sustain progress without constant intervention. In AI transformations, this long-term orientation is often the clearest signal that change will endure and real value will be delivered.

As traditional enterprise systems give way to agent-driven workflows, leaders must rethink how to structure teams and deliver work.

So what do customer support chatbots actually look like in the real world?

First – forget the “starting at $49/month” pricing you see on SaaS landing pages. That gets you a widget on your website. It doesn’t get you a system that actually resolves customer issues.

As hundreds of schools implement an automated monitoring tool, educators say that students can find talking to a chatbot ‘more natural’ than confiding in a human

You know your students. You know your subject. Maestro handles the complexity of putting the two together — so you can focus on what you actually do in the room.

I’m a cheapskate
I learn by failing
I love putting these elegant puzzles together
Friction is a clue something needs to be redesigned
Coding is my trauma Tetris
I’m afraid of making stuff that looks like everybody else’s
…
Some solutions need brooding
Friction sometimes helps me realize the idea is actually bad

Popular SaaS categories for investors now include startups building AI-native infrastructure, vertical SaaS with proprietary data, systems of action (those helping users complete tasks), and platforms deeply embedded in mission-critical workflows, according to Aaron Holiday, a managing partner at 645 Ventures.

But he also gave a list of companies that are considered quite boring to investors these days: startups building thin workflow layers, generic horizontal tools, light product management, and surface-level analytics — basically, anything an AI agent can now do.

Your conversations can be translated in real time, just by calling from any phone on our network. No apps, and no fees. Your voice. 50+ languages. Just talk and connect.

Instead of assuming that service or org transformation is always the answer, it’s worth pausing to explore which of these levels would provide the most value for your organisation, and whether you have the conditions in place to make change happen at the relevant level.