Anthropic has been so focused on knowledge worker use cases of AI. Avoiding the temptation of image and video gen, voice mode lags behind other labs, seemed limiting for a long time but it's been paying off big.

Product teams should take note of the compounding effect of focused discipline.

open.substack.com/pub/kylematt...

open.substack.com/pub/kylematt...

AI tools provide an instant "sugar high" of audio visual gratification, but lack the deep nourishment that comes from human struggle and intent. A compelling look at why the novelty of on demand art wears off so fast, and why the real artists are safe. open.substack.com/pub/kylematt...

Agents aren’t a “bigger model” problem. They’re an identity, guardrails, and orchestration problem. Part 3 of The Agentic Enterprise is live: kylematthies.substack.com/p/the-agenti...

Why docs fail: Good docs are time consuming to develop and maintain. If the 'why' isn't captured, new teams re-learn old lessons. People don't update docs. LLMs draft/refactor, but only if you have structured truth + metadata. Full playbook 👇

Quick guide to RAG platforms:
LlamaIndex — prototype fast, full control
LangChain — multi-agent orchestration
Pinecone/Weaviate — high-volume vector storage
Azure/Vertex RAG — regulated enterprise
Elastic hybrid — keyword + vector search

kylematthies.substack.com/p/part-2-fro...

"We already have docs" ≠ ready for agents.

Real process = spaghetti:
• Approval loops
• Threshold triggers
• Friday cutoffs
• The "ask Sam" step

If your map doesn't show where humans MUST approve, it's not real.

Agents need workflow truth, not fantasy.

open.substack.com/pub/kylematt...

Your agents are only as good as your documentation.

If the "why" behind decisions never got captured, if your process maps don't show approval loops, if pages are 5000 words of everything — your agents will hallucinate, skip steps, and sound incompetent. open.substack.com/pub/kylematt...

Copilots assist. RPA repeats. Agents think.
Part 1 of The Agentic Enterprise breaks down how to tell them apart—and where to actually trust autonomy.
open.substack.com/pub/kylematt...

AI workslop is real but the problem is bad management. LLMs aren’t magic tools, they’re junior teammates. ICs have to step into the manager’s chair: set outcomes, decompose tasks, write briefs, coach, fact-check, and package. That’s a skill gap.

open.substack.com/pub/kylematt...

If it were easy, the gurus would stop selling courses. The Solopreneur Trap → medium.com/@kyle.matthi... #ai #startups

RONALD REAGAN (1988): We should be aware of the demagogues who are ready to declare a trade war against our friends, weakening our economy, our national security, and the entire free world—all while cynically waving the American flag.

One of my party facts is that according to polls, Americans believe foreign aid is about 25% of the federal budget and on average that it *should* be about 10%.

The actual number is less than 1%.

Foreign aid has long been a boogey man of the right, so it's not surprising they're vilifying it now.