Archive
Every published post, newest first.
2026
Agentic AI Success Stories - Part 3 - Coding with human review
Part 3 of 3 — Agentic AI Success Stories The pull request is the product. Part 2 put a human at the start, before anything was written. Coding agents put a human at the end: merge. Where agents…
Click to continue readingAgentic AI Success Stories - Part 2 - Operations with a copilot
Part 2 of 3 — Agentic AI Success Stories Someone still typed the booking ID. Part 1 was support that talks to customers. This part is back-office work that writes — refunds, journals, tickets — and…
Click to continue readingAgentic AI Success Stories - Part 1 - Support that takes action
Part 1 of 3 — Agentic AI Success Stories Check the date of each number, then ask what happened next. This part uses those five questions. Support is where roundups get loudest: refunds, returns,…
Click to continue readingAgentic AI Success Stories - Introduction
An announcement doesn't prove it works in real use. Roundup articles will sell you eight company names and one big number. Where agents actually make sense already gave the test: a goal you can…
Click to continue readingMulti-agent systems are a graph - Part 4 - Where it shows up
Part 4 of 4 — Multi-agent systems are a graph Two kinds of graph. The domain graph is the set of links that already exist in your data: accounts to devices, services to deploys, suppliers to SKUs,…
Click to continue readingMulti-agent systems are a graph - Part 3 - When not to split
Part 3 of 4 — Multi-agent systems are a graph The cheapest graph is one node. Part 2 made fan-out look powerful. It is also expensive: more prompts, more traces, more places that need approval, and…
Click to continue readingMulti-agent systems are a graph - Part 2 - Fan-out, then merge
Part 2 of 4 — Multi-agent systems are a graph A path cannot do two jobs at once. Part 1 left you a map. A chain moves through it one step at a time. When work needs two specialists at the same time,…
Click to continue readingMulti-agent systems are a graph - Part 1 - The map
Part 1 of 4 — Multi-agent systems are a graph Adding agent names does not give you a map. The introduction said a multi-agent system is a graph. This part only defines its parts. What’s coming next…
Click to continue readingMulti-agent systems are a graph - Introduction
Adding more AI personas does not make a system. What’s coming next already named the linear chain: researcher → writer → reviewer. Orchestration decides when each step runs. Memory is where results…
Click to continue readingWhy AI Agents Fail - Part 5 - Knowing When — and What Reliable Agents Need
Part 5 of 5 — Why AI Agents Fail A truly capable agent is not one that always acts. Part 4 was about human over-trust. This part flips the definition of smart, then names the runtime that makes that…
Click to continue readingWhy AI Agents Fail - Part 4 - The Human Failure Mode
Part 4 of 5 — Why AI Agents Fail The agent is not the only unreliable component. Part 2 multiplied step risk. Humans multiply trust. Automation bias, fading expertise, and fuzzy ownership turn a…
Click to continue readingWhy AI Agents Fail - Part 3 - The Benchmark Illusion
Part 3 of 5 — Why AI Agents Fail One leaderboard is not a reliability certificate. Benchmarks are useful. They are also easy to misuse. Teams read a high score as “ready for production.” Different…
Click to continue readingWhy AI Agents Fail - Part 2 - The Seven Ways Agents Fail
Part 2 of 5 — Why AI Agents Fail Seven ways agents fail. Part 1 was about expectations. Enterprise war stories mix three kinds of failure: 1. Strategic / governance — no clear purpose or ROI; soft…
Click to continue readingWhy AI Agents Fail - Part 1 - The Illusion of Intelligence
Part 1 of 5 — Why AI Agents Fail A fluent answer is not necessarily a finished job. Demos reward the wrong signal. The model narrates a plan, names tools, and sounds sure. Viewers infer: if it can…
Click to continue readingWhy AI Agents Fail - Introduction - The Reliability Problem
What happens when an AI that is 95% right is given permission to act 100 times? Treat the numbers as illustration, not a lab measurement. A model that is usually right in chat still faces a different…
Click to continue readingRAG Explained - Part 5 - RAG vs Agents vs Automation
Part 5 of 5 — RAG Explained The wrong question is “which AI should we buy?” The right question is “which loop fits the job?” Part 1 argued for an approved corpus. Parts 2–4 showed the pipeline and…
Click to continue readingRAG Explained - Part 4 - Product Instances
Part 4 of 5 — RAG Explained These products are versions of the same pipeline. Part 2 drew index → retrieve → generate. Running that pipeline inside a company takes engineers, ACLs (Access Control…
Click to continue readingRAG Explained - Part 3 - Enterprise Ops
Part 3 of 5 — RAG Explained The demo indexed ten PDFs. Production indexes ten thousand — with ACLs. Part 2 drew the pipeline. Enterprise RAG is what happens when that pipeline meets real permissions,…
Click to continue readingRAG Explained - Part 2 - Architecture
Part 2 of 5 — RAG Explained RAG is a pipeline. Retrieval-augmented generation (Lewis et al., 2020) named the pattern: index a corpus, retrieve relevant passages, condition generation on those…
Click to continue readingRAG Explained - Part 1 - Why Your Data
Part 1 of 5 — RAG Explained The model passed the interview. It failed the handbook. General-purpose AI is trained on public text up to a cutoff. It does not hold your employee handbook, your ticket…
Click to continue readingRAG Explained - Introduction
General AI knows language. It does not know your policy PDF. That gap is why enterprise teams stop at the demo. The model answers fluently about refunds, retention, or last quarter’s outage, and…
Click to continue readingHow AI Agents Actually Work - Part 8 - Evaluation
If you only grade the last outcome, you did not grade the agent. Orchestration produces a path: model calls, tool calls, retrievals, stops. Evaluation is the layer that scores that path. It sits on…
Click to continue readingHow AI Agents Actually Work - Part 7 - Orchestration
Someone has to be the clock. The LLM proposes. Memory stores. Planning revises. Tools execute. RAG fetches. Orchestration is the runtime that sequences those moves: who is called, in what order, with…
Click to continue readingHow AI Agents Actually Work - Part 6 - RAG
Retrieval is a library card. Retrieval-augmented generation (Lewis et al., 2020) named a simple pattern: fetch documents, then generate. You index a corpus you approve. At ask-time you retrieve a few…
Click to continue readingHow AI Agents Actually Work - Part 5 - Function calling
The model fills out a form. Your code submits it. Part 4 listed the tools. Function calling (vendors also say tool use) is the protocol: the model does not “run Python in its head.” It emits a…
Click to continue readingHow AI Agents Actually Work - Part 4 - Tools
Tools are the actions an agent can take. The agent era already covered the history: chat got APIs, browsers, code, apps. A tool is a capability your runtime can actually run, with a name, a purpose,…
Click to continue readingHow AI Agents Actually Work - Part 3 - Planning
A plan that cannot change after contact with the world is a script. Part 2 covered memory (the notebook). Planning is what you do with it: turn a goal into steps, then re-plan when an observation…
Click to continue readingHow AI Agents Actually Work - Part 2 - Memory
If nothing is written down, the next call starts amnesiac. Part 1 left the model as a stateless engine: context in, tokens out. Memory is the architectural answer to “what do we put in that context…
Click to continue readingHow AI Agents Actually Work - Part 1 - The language model
The model is an engine. An agent is a vehicle. A large language model predicts plausible next tokens given the text so far. That is the whole magic trick, and it is enough to write fluent paragraphs.…
Click to continue readingHow AI Agents Actually Work - Introduction
An agent is not a smarter chatbot. It is a stack. From Models to Agents maps the eras: chat, reasoning, tools, then the enterprise approach. AI Right Now Part 6 covers control: what an agent may…
Click to continue readingFrom Models to Agents - Part 10 - What’s Coming Next
Near-term agents get more tools. Parts 1–9 mapped chat → reasoning → tools → loop → risk → fit → policy → eval → ship. This last part names patterns already peeking through. None of them require…
Click to continue readingFrom Models to Agents - Part 9 - Shipping and Deployment
Productising an agent is operations. Prompt engineering and policy said what it may be told and what it may touch. Evals said it behaves on the cases you care about. Shipping is how it lives:…
Click to continue readingFrom Models to Agents - Part 8 - Evaluation and Testing
Most production agents are not “trained.” They are evaluated until they behave. Part 7 wrote the specification. This part is about measuring the agent, again and again. Shipping waits for Part 9. If…
Click to continue readingFrom Models to Agents - Part 7 - Prompt Engineering and Policy
Writing the agent is writing the prompt and the policy. Part 6 asked where agents fit. This part is design time: two documents that have to agree. The prompt is what the model is told (goal, scope,…
Click to continue readingFrom Models to Agents - Part 6 - Where Agents Actually Make Sense
Start where “done” is checkable and undo is possible. Part 5 was the cold water. This part is the filter. Agents are not a personality upgrade for every workflow. They pay off where a goal can be…
Click to continue readingFrom Models to Agents - Part 5 - The Uncomfortable Reality
Agents can still make mistakes — fluent, confident mistakes. The loop is powerful because it compounds. That is also why it is uncomfortable. A wrong token in chat is easy to ignore. A wrong token…
Click to continue readingFrom Models to Agents - Part 4 - Why Agents Are Different
Chat is a conversation. Agents are a control loop. Part 3 added tools. This part is why that is not “chat with plugins.” The way the system works changes. An error no longer stays in one paragraph.…
Click to continue readingFrom Models to Agents - Part 3 - The Agent Era
Agents keep the language model, then add tools. Chat replies. Reasoning plans more carefully. An agent tries to finish a job: read a database, call an API, browse the web, run code, click through an…
Click to continue readingFrom Models to Agents - Part 2 - The Reasoning Era
The next leap was not a friendlier chat box. It was more thinking time. In Part 1 the loop was prompt → fluent reply. That is fast, and it is often shallow on hard problems. The reasoning era is what…
Click to continue readingFrom Models to Agents - Part 1 - The Chatbot Era
We spent years making AI better at answering questions. That habit still shapes how most people use the technology. You type. It replies. You edit. If it goes wrong, the damage is mostly words on a…
Click to continue readingFrom Models to Agents - Introduction
We’ve spent years making AI better at answering questions. What happens when we start asking AI to actually do things? That is the shift to watch, more than the next model name, more than the next…
Click to continue readingAI Right Now - Part 14 - How to Live Well With AI
You don’t need to become an engineer. You need a personal AI policy. Fourteen pages ago we started with a quiet truth: AI is already in your morning. Along the way we named the toolbox, demystified…
Click to continue readingAI Right Now - Part 13 - What Comes Next (Near Term)
The future is closer — and narrower — than sci-fi suggests. Sci-fi loves overnight robot overlords. Rational planning loves near-term trends you can already see: better tool-using agents, cheaper and…
Click to continue readingAI Right Now - Part 12 - AI at Work: Real Gains, Real Friction
Companies bought AI. Culture did not install itself. Licenses are easy to purchase. Habits are not. The gap between “we have ChatGPT” and “our work actually improved” is where most of the quiet…
Click to continue readingAI Right Now - Part 11 - Safety, Trust, and the Rules Catching Up
The speed of invention is outrunning the speed of trust. New demos arrive weekly. Scams arrive weekly too. Deepfakes, phishing with perfect grammar, privacy leaks, over-confident advice — the same…
Click to continue readingAI Right Now - Part 10 - Money, Power, and the AI Race
Behind every “free” AI chat is a very expensive engine. Electricity. Chips. Data centers. Specialist talent. Safety teams. Legal reviews. Distribution platforms that already own your attention. When…
Click to continue readingAI Right Now - Part 9 - Jobs, Skills, and the New Division of Labor
AI is not coming for “all jobs.” It is coming for tasks inside jobs. That sentence will disappoint two groups: people who want a simple apocalypse story, and people who want a simple “nothing will…
Click to continue readingAI Right Now - Part 8 - Where AI Still Fails
The most dangerous AI answer sounds perfect — and is wrong. A clumsy error is easy to catch. A smooth error slides into your essay, your slide deck, your customer reply. Confidence can hide…
Click to continue readingAI Right Now - Part 7 - Where AI Wins Today
Ignore the hype. Look at the results. Movies promise robot bosses and overnight utopia. Real life offers quieter wins: faster drafts, clearer search, translation that unblocks a conversation, code…
Click to continue readingAI Right Now - Part 6 - Agents: When AI Starts Doing Work
The next leap is not smarter chat — it’s AI that takes steps. Chat answers. Agents attempt to finish a job. That sounds small. It is not. Advice is cheap. Action has consequences: money moved, emails…
Click to continue readingAI Right Now - Part 5 - Multimodal AI: Seeing, Hearing, Speaking
AI stopped being only a chat box. For a while, the public face of AI was a chat window. Type words. Get words. Useful — and incomplete. The next wave connects language with images, audio, and…
Click to continue readingAI Right Now - Part 4 - The Rise of Large Language Models
One idea changed the internet: machines that talk like us. Not perfectly. Not always truthfully. But fluently enough that the old “press 1 for sales” chatbot suddenly felt ancient. Large language…
Click to continue readingAI Right Now - Part 3 - How Modern AI Learns (Without Magic)
AI doesn’t “know.” It gets very good at guessing. That sentence can sound cold. It is meant to be freeing. Much of modern AI is pattern learning: look at huge piles of examples, notice what tends to…
Click to continue readingAI Right Now - Part 2 - Not One AI — A Whole Toolbox
Saying “AI” is like saying “vehicles.” Cars, bikes, and planes all move people. Nobody treats them as the same machine. Yet we often say “AI” as if it were one product, one mind, one future. That…
Click to continue readingAI Right Now - Part 1 - Introduction
You used AI today. Maybe you did not open a chatbot. Maybe you never typed the letters “A” and “I” into a search bar. It does not matter. If you unlocked your phone with your face, checked a map for…
Click to continue readingWelcome to Liliputtech.io
Welcome. I’m glad you’re here. Hi. I’m Sakti, and I write about technology the way I wish someone had explained it to me years ago: clearly, honestly, and without the hype. I am not here to impress…
Click to continue reading