Permalink: 20260901175620

Found at “Topological Sorting using BFS - Kahn’s Algorithm” on 2026-09-01T17:56:20+02:00.

Given a Directed Acyclic Graph having V vertices and E edges, find any Topological Sorted ordering of the graph.

Permalink: 20260831164641

Found at “From Data and APIs to Capabilities: The Abstraction the Agentic Era Is Settling On” on 2026-08-31T16:46:41+02:00.

In June 2023 — before MCP existed, before anyone was seriously designing for agents — Bruno Pedro asked what the unit of value in an API marketplace actually is:

“I believe the unit of value in APIs is a single capability and not the whole API… Consumers want the capabilities provided by the APIs, not the APIs themselves.”

He noticed something that reads differently now: every consumer platform had already invented its own word for the same object. Google and Apple called them Actions, Amazon Skills, ChatGPT Plugins. Four vendors, four words, one unit — and nobody was designing for agents. They were designing for consumers who wanted to do a thing, and the API was in the way.

Permalink: 20260827142011

Found at “semantic-conventions-genai/docs/gen-ai/gen-ai-spans.md at main” on 2026-08-27T14:20:11+02:00.

GenAI spans represent logical operations as observed by the caller. They SHOULD cover the duration of the operation, starting when it is initiated and ending when the response is fully received or the operation is terminated due to an error or cancellation.

Permalink: 20260816161958

Found at “Which AI Models Watermark Their Output? (2026)” on 2026-08-16T16:19:58+02:00.

Several major AI models now watermark their output, but coverage varies widely by provider and by content type. As of August 2026, Claude is the clearest confirmed example of an embedded, machine-readable text watermark applied with no opt-out. Google’s SynthID also watermarks generated text inside Gemini. Most other providers mark images and audio today, and their text-watermarking status is either newer or not publicly confirmed.

Permalink: 20260814144813

Found at “How Claude marks AI-generated content” on 2026-08-14T14:48:13+02:00.

As AI-generated content becomes commonplace, greater transparency and signals about where content comes from can give people useful context about the information they consume. To support transparency and comply with our legal obligations, Anthropic is working to include machine-readable marks in content that Claude generates.

Permalink: 20260814144544

Found at “Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Models” on 2026-08-14T14:45:44+02:00.

Watermarking generative models consists of planting a statistical signal (watermark) in a model’s output so that it can be later verified that the output was generated by the given model. A strong watermarking scheme satisfies the property that a computationally bounded attacker cannot erase the watermark without causing significant quality degradation.

Permalink: 20260723120822

Found at “From Capabilities to Responsibilities” on 2026-07-23T12:08:22+02:00.

High-stakes AI systems must be designed around responsibilities, not capabilities.

Permalink: 20260723120702

Found at “Industrialism and the slippery slope” on 2026-07-23T12:07:02+02:00.

Quality, by definition, is meeting spec. If you don’t like the spec, make the spec better.

Permalink: 20260719105922

Found at “The New Software Lifecycle” on 2026-07-19T10:59:22+02:00.

AI compresses the lifecycle, but unevenly, and the unevenness is the whole story. Implementation drops from weeks to hours. Requirements, architecture, and verification stay slow because they’re judgment work. So specification quality becomes the bottleneck, and verification moves to the middle.—Addy Osmani

About this new AI-driven SDLC, Kin Lane asked me and other API product folks on LinkedIn:

How does this apply to the API lifecycle?

This is my answer:

The article and the paper it refers to validate the importance of doing API Design the right way: capturing JTBD, writing user stories, understanding consumers’ needs, iterating using API prototypes, and validating before implementing.

Building things incorrectly isn’t a risk anymore because the cost of implementation has dropped dramatically. LLMs can now build what you ask them to in any programming language, bug-free. What’s now a risk is building the wrong things. That’s why validating your API design is so important.

A “side effect” of this new SDLC is that now almost anyone can build an API. You don’t need to be a developer to be able to design an API, implement it, and release it. You don’t even have to know OpenAPI or any other API specification. What you do need to know, however, is your domain. You need to understand what you’re building and who you’re building for.

Permalink: SDD for API Design

Today, I put together a few pieces that looked totally disconnected until now.

I was reading some random article about how AI, and in particular the specification-driven development (SDD) methodology, can open software development to non-technical people. The first step of SDD is to define the specification of the software that you want to develop.(1) (1)“This isn’t about technical stacks or app design. It’s about user journeys, experiences, and what success looks like,” wrote Den Delimarsky in September 2025 (see Spec-driven development with AI: Get started with a new open source toolkit). In October 2025, Birgitta Böckeler added that “a spec is a structured, behavior-oriented artifact (…) written in natural language that expresses software functionality” (see Understanding Spec-Driven-Development: Kiro, spec-kit, and Tessl). Who has better knowledge of what should be built than someone who has expertise in their area of work?

I immediately thought about the kinds of profiles that thrive in an AI-oriented world and job market. Expert generalists, or T-shaped professionals, are probably going to rule this AI wave.(2) (2)“We’ve observed that Expert Generalist capabilities are considerably more valuable with these LLMs,” wrote Martin Fowler et. al. in July 2025 (see Expert Generalists). There’s no room or time for specialists. Why? Because writing code is cheap.(3) (3)“Coding agents dramatically drop the cost of typing code into the computer,” wrote Simon Willison in February 2026 (see Writing code is cheap now). Translating is cheap. Building things is cheap. You know what isn’t cheap? Understanding what to build and understanding why you should build it.(4) (4)“The ability to define what needs to exist is becoming the scarcest and most valuable skill in any organization,” wrote Ivan Misic in May 2026 (see When AI Makes Building Cheap, Knowing What to Build Becomes Everything). That’s something an expert generalist can do better than a specialist.

And then I thought that SDD can be used by expert generalists to design and build APIs. APIs that really make sense and offer value to consumers. Expert generalists have the knowledge of their area of expertise and are generalist enough to understand what API operations they should build.

Permalink: 20260623105720

By classifying operations as atomic, workflow, asynchronous, introspective, discovery, reactive, bulk, or simulated, you create a predictable map that helps machines navigate APIs safely.—The Taxonomy of API Operations

Inferred API operation categories. This is the distribution of API operation categories after analyzing more than 50k OpenAPI documents from public GitHub repositories. I inferred a taxonomy by comparing the morphology of each API operation with a set of pre-established heuristics.

Permalink: 20260617111326

Found at “Expert Generalists” on 2026-06-17T11:13:26+02:00.

We’ve long noticed that not just anyone succeeds as an Expert Generalist, but once we understand the traits that are key for such Expert Generalists, organizations can shape learning programs, hiring filters, and career paths that deliberately develop them.

Permalink: 20260604110757

Found at “Enshittification, Despotification, and the Open Internet” on 2026-06-04T11:07:57+02:00.

There’s no reason we can’t build a new generation of services that restores that democratic, decentralized promise—an open internet that empowers users rather than funneling control to gatekeepers.

Permalink: 20260526104220

Found at “Appearing Productive in The Workplace” on 2026-05-26T10:42:20+02:00.

Generative AI can produce work that looks expert without being expert, and the failure arrives in two shapes. The first is when novices in a field are able to produce work that resembles what their seniors produce, faster or more advanced than their judgment. The second is when people generate artifacts in disciplines they were never trained in. The two failures look similar from a distance and are not the same. Research has mostly measured the first. The second is what it is missing, and in my experience it is the riskier of the two.

Permalink: 20260525123455

Found at “From Hierarchy to Intelligence” on 2026-05-25T12:34:55+02:00.

At Block, we’re questioning the underlying assumption: that organizations have to be hierarchically organized with humans as the coordination mechanism. Instead, we intend to replace what the hierarchy does. Most companies using AI today are giving everyone a copilot, which makes the existing structure work slightly better without changing it. We’re after something different: a company built as an intelligence (or mini-AGI).

Permalink: 20260505171152

I’m currently starting a new line of research about API-related taxonomy. Here’s a very early exploration on the taxonomy of API operations, where I identify eight operation categories:

• Atomic
• Workflow
• Asynchronous
• Introspective
• Discovery
• Reactive
• Bulk
• Simulated

This is very early work, which means that even this list of categories isn’t yet final.

Permalink: 20260505102545

Found at “Agent Skills” on 2026-05-05T10:25:45+02:00.

AI coding agents are extremely capable junior engineers with no instinct for the parts of the job that don’t show up in the diff. The senior-engineering work (surfacing assumptions, sizing changes, writing the spec, leaving evidence, refusing to merge what can’t be reviewed) is exactly what an agent will skip unless you make it impossible to skip. The job, increasingly, is to encode that discipline as something the agent cannot talk itself out of.