Author | Lightnews

Schwentker @schwentker.bsky.social · 16h

.@paddix CEO of @digitalocean - “When you have fresh, relevant data delivered at very low latency, you gain relevancy.”

The shift isn’t bigger models or better search.

It’s search embedded in everything, alive in real time.

1

Schwentker @schwentker.bsky.social · 16h

If relevance now decays in minutes, what replaces attention as the web’s true currency?

The answer, argues "Half-Life Problem" is continuity --data that regenerates instead of rots.

linkedin.com/pulse/half-l...

Half-Life Problem: Why Search Indices Became Archaeology

Inside Web Discovery Summit 2025, Or Lenchner and Paddy Srinivasan reveal the temporal crisis forcing search architecture to rebuild from scratch Executive Summary: Or Lenchner Bright Data and Paddy S...

linkedin.com

1

Schwentker @schwentker.bsky.social · 8d

If AI agents can buy, who do they trust?

If tokens self-destruct, what anchors value?

Agentic Commerce Protocol (ACP) turns payment from final click to ongoing negotiation.

Secure Payment Tokens (SPT): sealed letters of credit that expire.

www.linkedin.com/pulse/when-a...

When Agents Shop: Inside PayPal's Architecture of Autonomous Commerce

When AI agents became shoppers, something fundamental shifted. Not the interface -- the topology of commerce itself.

www.linkedin.com

1

Schwentker @schwentker.bsky.social · 9d

Stateless systems scale by ignoring the past.

Stateful agents scale by remembering it.

@charlespacker:
“The next advancement in AI won’t come from bigger models, but from agents that learn from experience.”
To move forward, first remember.

The paradox: to move forward you must first remember.

Schwentker @schwentker.bsky.social · 9d

When building AI agents, what should you retain in memory - & what should you forget?

As @venks from @Intuit put it: “We’re using LLM-as-a-Judge to measure agent’s accuracy & usefulness.”

How will you version the memory your agent carries forward?

#AIagents #observability #ops

1 2

Schwentker @schwentker.bsky.social · 9d

“Memory is more than context windows” — @charlespacker on stateful agents & long-horizon reasoning.

Check out my recap of the Intuit meetup: “Stateful Futures: @Letta_AI, @Intuit & Architecture of Perpetual Agents”

→ linkedin.com/pulse/statef...
#AI #agents #infrastructure

Stateful Futures: Letta, Intuit & Architecture of Perpetual Agents

A decade of cloud computing trained organizations to think stateless: spin up, execute, discard. Now the grammar inverts.

linkedin.com

1 1 3

Schwentker @schwentker.bsky.social · 15d

Tonight in San Francisco

QR-coded greenroom door @Vapi_AI set the tone: automation w/o awareness.

Agentic = behavior, not chrome.

Generation outruns comprehension.

Execution automates, taste does not.

w/ @juliusai @magicpatterns @withdelphi

www.linkedin.com/pulse/design...

Design After Design: Field Notes from Interface Frontier

Tonight the powder room door at Vapi's headquarters unlocks via QR code: a micro-symptom of the evening's theme. Systems that predict your needs while forgetting you might need to leave the room.

www.linkedin.com

1

Schwentker @schwentker.bsky.social · 16d

.@beyarkay Boyd Kane’s essay on why AI can’t be “debugged” like software is lighting up @hackernews -

engineers & executives suddenly realizing they’ve been seeing the same system through incompatible lenses.

The overlooked risk in AI deployment ↓
linkedin.com/feed/update/...

Why Boyd Kane's AI vs Software Essay Resonates & the Overlooked Risk Hidden in Deployment Decisions Watched an essay cross 260+ upvotes yesterday with a tone shift worth noting: not hype fati...

Why Boyd Kane's AI vs Software Essay Resonates & the Overlooked Risk Hidden in Deployment Decisions Watched an essay cross 260+ upvotes yesterday with a tone shift worth noting: not hype fatigue or d...

linkedin.com

1

Schwentker @schwentker.bsky.social · 17d

The agent sits between buyer & business, controlling checkout UI & holding payment tokens.

When software intermediates commerce at this level, where does the merchant relationship actually reside? With who renders the interface?

Schwentker @schwentker.bsky.social · 17d

6-month prediction: w/ Shopify's 1M+ merchants entering test mode for ChatGPT Instant Checkout, we'll see meaningful uptake by Q1 2026.

Protocol adoption follows a pattern: early infrastructure layer (done), reference implementation (live w/ Etsy), then the land grab (Shopify rollout = this).

1

Schwentker @schwentker.bsky.social · 17d

When agents spend on behalf of humans, where does intent reside?

Stripe & OpenAI just architected the payment rail for AI commerce.

But the protocol reveals something deeper:

we're building infrastructure faster than we understand its implications.

linkedin.com/pulse/protoc...

Protocols, Not Platforms: Tacit Architecture Behind Agentic Commerce

How Stripe & OpenAI Just Invented the Payment Rail for AI Agents In September 2025, two tech giants architected a solution to a problem most didn't yet recognize: how AI agents transact on behalf of h...

linkedin.com

1

Schwentker @schwentker.bsky.social · 23d

Prediction: Agentic frameworks (LangGraph, CrewAI, LlamaIndex, n8n) will adjust strategy significantly in next 6 months as both OpenAI's AgentKit & Anthropic's Agent SDK reshape market.

Not extinction - evolution. Specialization deepens where platforms can't follow. Stay tuned.

Schwentker @schwentker.bsky.social · 23d

Question worth considering: If AgentKit commoditizes orchestration complexity where does differentiation migrate next?

To eval quality? Specialized reasoning? Domain-specific tool libraries?

Layer that remains hard may determine which frameworks capture value over next 24 months.

1

Schwentker @schwentker.bsky.social · 23d

OpenAI DevDay 2025 shifted the conversation from "do agents work?" to "who controls the infrastructure they run on?"

New analysis on AgentKit vs LangGraph, CrewAI, LlamaIndex & n8n - comparing not just features but philosophies about where complexity should live.

www.linkedin.com/pulse/openai...

OpenAI DevDay 2025: Agents Reach Operational Reality

At OpenAI DevDay 2025, the agent development conversation shifted terrain - from whether agents work to who controls the infrastructure layer that makes them work at scale. The question isn't new.

www.linkedin.com

1

Schwentker @schwentker.bsky.social · 26d

"The model already has a lot of capabilities. I'm convinced that even w/ current generation models, there's way more intelligence in there than we've been able to unlock."

@brada, Product Lead, Claude Developer Platform

on unhobbling AI vs constraining it w/ scaffolding

Schwentker @schwentker.bsky.social · 26d

"We can help you get self-improving & continuously improving outcomes out of Claude. Not better execution of predetermined plans. Outcomes that improve themselves through intelligent iteration."

Claude's @katelyn_lesse on what happens when frameworks become liabilities

1 1

Schwentker @schwentker.bsky.social · 26d

What processes assume software takes weeks to build?

If intelligence compounds monthly what does annual planning mean?

When software gens itself what becomes the product?

3 questions as autonomous agents build Slack in 30hrs & replicate APIs in 5hrs

www.linkedin.com/pulse/softwa...

Software Development Has Already Changed. Most Haven't Noticed.

Three data points arrived within weeks of each other. Each one unremarkable on its own.

www.linkedin.com

1