Welcome, humans.

We now have more insight into what OpenAI’s upcoming software engineer will do:

Called “A-SWE” (a name that OpenAI CFO Sarah Friar absolutely drags in the above clip), the agent builds entire apps for you, handling coding, testing, bug fixes, and even creating the documentation developers hate writing but everyone needs to understand and maintain your code.

Sounds awesome, right? But in reality, many code bases are messy, and making edits that account for all the ripple effects is quite challenging.

For example, AI agents right now often hallucinate software packages (pre-written plug and play code) that doesn’t exist.

On top of that, hackers have now realized they can exploit this issue to create malicious software packages under the fake names.

But hey, at least when the robot uprising happens, they'll have perfectly documented code explaining exactly how they took over. Take that, human devs!

Here’s what you need to know about AI today:

• We break down how to automate anything.

• Ireland investigated X.com for using personal data to train Grok AI.

• OpenAI will replace GPT-4 with GPT-4o in ChatGPT on April 30.

• Ex-OpenAI staff supported Musk's lawsuit against OpenAI's for-profit shift.

How to automate (almost) anything with AI.

This is the year of the agents, they said. Everyone needs to add agentic workflows, they said. Yeah, yeah, but what does that actually mean?!

There’s lots of definitions of agents out, but the one that makes the MOST sense to us (from Latent Space) is this:

Let’s break that down:

1. Language model = an AI like ChatGPT. You input language tokens (AI speak for words turned into numbers), you get language tokens output back ‘atcha.

2. Memory = the context window, or the ability to hold lots of information in its attention span without getting confused.

3. Planning skills = “thinking mode”, which is basically just the AI running through scenarios until it lands on the most likely solution.

4. Tool use = the ability to use functions, like code that’s already written to accomplish a specific task (just like we humans do when WE use computers).

For example, the popular trend now is making “browser use” agents like OpenAI’s Operator, China’s Manus, or the appropriately named “Browser Use” that can navigate the web like humans can (there’s also LMNR, which we tested this weekend).

These agents use a language model to plan out a task (whatever you ask it to do), use the browser as a tool, and update their memory as they progress towards that goal. Claude plays Pokemon works the same way.

BTW, the most powerful agents use multiple language models doing different roles to accomplish different tasks in a workflow:

1. Say you’ve got your planning AI…

2. Who passes to your writer AI…

3. Who passes to your editor AI…

4. Who maybe passes to a fact checker AI…

5. And then finally passes to your publisher AI.

Put all those together, and you’ve essentially got a “publishing agent” that can publish complete works for you.

So how do you make your own agent? A lot of companies (and we mean, A LOT) offer “agents” that are some combination of all of the above. Not all of these are real agents, though.

What makes a true agent different from an “agentic workflow” (fancy talk for putting a series of automated steps together that involve AIs) is that an agent does the reasoning and planning for you—so instead of you deciding the best way to do everything, the AI does (Jeff Su explains this well).

Now, in order to make your OWN agent (to automate anything), we’ll focus first on the “original” version of an agent (the agentic workflow) so you can practice thinking through everything you need to tell your AI to do for you.

The two main tools non-coders (and some coders) use to do this are Make and n8N. Make is the every man’s tool, while n8n is the coder’s preference (it’s kinda like the Claude to Make.com’s ChatGPT—all the cool kids use n8n).

We recommend learning n8n, but it can feel really complicated, like, as soon as you start. Luckily, Tina Huang put together this epic tutorial, and n8n has their own tutorial series, too.

That said, if Make is more your style, this tutorial from Helena Liu is the one for you.

Those three videos alone should be enough to get you started, but we put together a whole explainer on the website with more resources, too—check it out here.

FROM OUR PARTNERS

Could RYSE be the next Ring?

Imagine skipping out on Ring before its $1.2B buyout by Amazon.

That’s exactly what happened to all the sharks, including Kevin O’Leary, missing a 66,756% return.

By the time we hear about industry-changing companies, it’s usually too late. But right now, there’s a smart home startup making their way to homes in America.

This tech startup is RYSE, and unlike Ring, you can still invest early—before it takes off.

Like how Ring disrupted home security, this company is revolutionizing smart blinds & shades—bringing automation to every home and business without the need for expensive replacements.

With $10M+ in revenue, sold in 127 Best Buy stores, and 10 patents, they are primed for massive expansion.

Shares are now available at $1.90 each in their newly launched funding round.

Learn more

Prompt Tip of the Day

Here’s a prompt for how to make extremely realistic looking images—friendly reminder to make these for fun, and not to distort everyone’s shared reality any further than it actually is.

Treats To Try.

1. Microsoft’s New Windows 11 build introduces AI features including Recall for searching past activities, Click to Do for taking on-screen actions, and semantic indexing for improved Windows Search—gonna need a Copilot+ PC to use it tho.

2. CapCut launched Pippit, a creative agent that transforms websites into marketing videos and generates avatars of you to streamline your marketing content production (video)—free 30 day trial rn.

3. You can now integrate HeyGen with HubSpot to create personalized videos for email campaigns without filming anything—paid starting @ $23.92/month (video).

4. Marketers are also using AI like Creatify, Arcads, or MakeUGC or to create user-generated content (UGC) style ads—friendly reminder to trust nothing you see online!

5. OpenCreator combines the 20+ top AI creative tools in one place so you can produce content without switching platforms or paying multiple subscriptions (here’s a tutorial from El.Cine on X)—free to try.

6. Higgsfield transforms your videos with one-click professional camera motions like crash zooms, explosions, and drone effects.

7. Someone used Gemini 2.5 to make an old-school Atari Missile Command clone.

See our top 51 AI Tools for Business here!

Around the Horn.

• Ireland opened an investigation into X.com over its use of users’ personal data to train its Grok AI model.

• OpenAI will remove GPT-4 from the ChatGPT model picker on April 30 and replace it with GPT-4o (it’ll still be available via API though).

• 12 Ex-OpenAI employees filed an amicus brief (a “friend of the court” brief from those not involved in the case) in favor of Elon Musk’s case against OpenAi becoming a for-profit.

• In another AI legal case, a group of law professors wrote their own amicus brief (as did the global publishers’ trade group) in support of authors suing Meta over copyright infringement.

FROM OUR PARTNERS

Why Consider Local GPU Compute?

Dell's “Reshaping Workflows” podcast shows professionals using Dell Pro Max workstations with NVIDIA RTX GPUs in real-world scenarios. First three episodes cover: the Dell Pro Max portfolio, NVIDIA's RTX Pro™ Blackwell GPUs for AI, and Dell's workstation redesign benefits.

Listen now. 

Sunday Funnies

A Cat's Commentary.