Notable verbatim quotes

⚠️ Quotes preserved verbatim, including transcription artifacts. Cite by section.

On the new paradigm

• Thesis (§1): "software development is transforming from revolving around code and instructions to revolving around intent… revolving around code and implementation to revolving around the intent and instructions."
• Stack-as-software (§6): "the tools, the harnesses, the factory length, the factories they're all actually software. They're software that wraps the model."
• Context is the new code (§6): "a lot of this is frameworks, but context is the new code. It is the place in which you program the model."
• Summary line (§11): "within those skills are the new code. And we should treat them that way and give them the right tools for it."

On tools

• (§3) "Tools are probably the easiest to understand. These are pieces of software, their utilities that in part indeed they turn models into agents."
• (§3) "It's not always raw intelligence that you want for every problem."

On context & skills

• (§4) "information that the agent either doesn't have and can't know or information that you can[t] figure out, but it's very inefficient [or] error[-prone] for it to find out."
• (§7) "skills are really designed to be reusable. They're light libraries… If you think you're going to do it multiple times, you turn it into [skills]."
• (§4) "composability is what makes software powerful… Imagine every time you want to write a web page, you have to write a kernel. Like that doesn't work."

On harnesses

• (§5) "harness is deterministic software that wraps a probabilistic model. It harnesses the model."
• (§5) "this commandeers control [from] the model. It says the model doesn't get to choose."

On the three risk types

• (§8) "malicious skills that are literally built to cause harm. And [in the] open floor ecosystem. We've seen over 30% of skills to be malicious."
• (§8) "negligence skills… skills that really urge the user to do something but do not have any safety instructions in it. Do not set any boundaries… make sure you update the table correctly… but do not drop the table."
• (§8) "vulnerable skills that guide for information [that] just leaves you exposed. Most common example of that is to use API tokens or other secrets as part of something that is visible in the logs."

On the shared-repo failure

• (§8) "There are seven over here. Which one should I choose? Like there's no indication of which one is good."
• (§8) "nobody trusted anything in the repo and eventually everybody came back to writing their own."

On lifecycle & rot

• (§8) "software rots. If you write a piece of excellent software today and you do not maintain it over time, it will stop working and eventually it will be harmful."
• (§8) "if you do maintain, you can actually turn maintenance into optimization."
• (§8) "both of them boil down to high quality autonomous agent powered maintenance."

On evals as tests

• (§9) "You take, you define an environment for some skill to run, you actually practically run an agent through a task in that environment. And then you judge the result."
• (§9) "scenario quality matters… you can write sounds [scenarios] that are just a waste of time tokens."

On dependencies

• (§9) "no one ever said dependency management. It's fun. Dependence management is a pa[in] and dependencies are pa[in]."
• (§9) "just like npm doesn't care if you're on windows or a [Linux] machine. It's just obvious… Can we do the same… for agents?"

On the lifecycle

• (§10) "it should be a context development life cycle not a skill, not a software… life cycle. We should as humans live in the context [development] life cycle and leave the SDLC to the agents."

On community

• (§12) "building a new paradigm is not a vendor or any company's job. It is a community activity."