knowledge-management on S Anand

Agent-consumable content

Tue, 19 May 2026 11:08:59 +0800

I’m making more and more of my content agent-consumable, i.e. easier for ChatGPT, Claude Code, etc. to read, in three ways.

One, I export content in an agent-friendly way.

Google email, calendar, chat. I use gws to back up into scannable one-line entries.
Meet recordings. I back up transcripts and videos (with a compact audio copy).
WhatsApp chats that I back up into similar one-liners.
Browsing history by exporting my Edge history SQLite database.
Daily activities by integrating the above with my command line and commit history.
AI conversations by exporting them manually or via bookmarklets.
Social media records like LinkedIn invites/conversations, Twitter, Hacker News, Discourse, etc via bookmarklets or scripts.
Financial records like bank statements, receipts, payslips, tax filings, utility payments, rentals, property records, investments, insurance, pensions, invoices, credit scores, etc. by exporting them manually.
Medical records like tests, prescriptions, doctor visits, etc. by exporting them manually.
Personal records like certificates, educational records, CV, passport / visa applications, etc. by exporting them manually.

Two, I log / generate more content. For example:

Things I learnt and blog posts I write.
Prompts I use frequently.
Trending GitHub repos I want to evaluate.
Talks I deliver and data stories I write.
Demos I build and code I write.
Weight and other fitness data.
Teaching material, assessments and evaluations and analysis.
Coding agent logs.
Sensors: Location, mostly.
Daily journals: Food, sleep, deeds, pains, …
Media journals: Books, movies, TV series, …
Notes (of various kinds): TODOs, app / tool ideas, people I know / meet, questions I’m asked, my beliefs, …

(Notably missing are photos / videos, which I’ve fallen out of the habit of.)

Three, I summarize the content for agents. For example:

Adding blog frontmatter by summarizing my blog posts
Adding transcript frontmatter by summarizing my meeting transcripts.
Identifying actions extracted from transcripts.
Summarizing my code as podcasts.
Summarizing prompts as SKILL.md files.
Summarizing conversations into advice for AI, time management, etc.
Summarizing technical choices into a technology radar
Extracting transcripts elements, like insights, experiments to run, actions, what I missed, what they missed, etc.

On my list is Karpathy’s LLM wiki, summarizing my photos, and more.

Just writing this post took me an hour! It also convinced me that I have lots of content and there’s a lot of under-leverage in unleashing agents on what I already have.

Organisational amnesia

Tue, 19 May 2009 20:23:36 +0000

It’s amazing how much of a dependency there is on individuals writing IT systems. Reminds me of that Dilbert strip:

A few weeks ago, I was trying to figure out in what happens when there are multiple promotions. (Our client is a retailer.) I mean, if there’s a phone that costs £100 and there are 2 promotions: 10% off on phones and £10 off on phones. Do you apply the 10% off first and pay £80 or the £10 off and pay £81?

Funnily enough, the organisational answer is, “I don’t know.” The person who determined the logic is no longer with the firm. The person who wrote the code was a contractor and moved on to another project. The vendor hadn’t gotten around to documenting the code. Sure, the code’s there, and you just had to read it to figure out what it does. But no human knew what it was supposed to do.

Last week, there was a decision to rewrite some code that was 10 years old. A colleague who wasn’t quite involved in this work said, “I’m going to have to set aside 2-3 weeks for this. I wrote this stuff when I was a developer. The docs have vanished. The business owners have vanished. I’m the only one who has any clue on what it’s supposed to do.”

This week, we were trying to figure out how their store locator system works. After fiddling around with Fiddler, and seeing that it used Microsoft Virtual Earth, I was able to figure out that it identified stores near a location using a simple JSON API. But can we get the documentation around that? Nope. Tough luck. Nobody knows how it works any longer.

Personally, I don’t think this is unusual. We forget. Companies forget. But it’s usually good if what we forget is derivable. That’s how I got through my high-school physics exams: not by remembering stuff, but by being able to derive the stuff from a few principles.

Organisations can do the same. But to be able to do that, you need to have commonly understood principles. As Fred Brooks put it in The Mythical Man Month,

I contend that conceptual integrity is the most important consideration in system design. It is better to have a system omit certain anomalous features and improvements, but to reflect one set of design ideas, than to have one that contains many good but independent and uncoordinated ideas.

One of the biggest enemies of conceptual integrity is growth. Too many people too soon, and the important decisions are taken by people who’ve never had a long chat about things. There’s another reason not to grow too fast.

Comments

Kalpesh 20 May 2009 2:27 am: I guess, it should help having a debug log of things as the code is executed & the decision made in the code based on the data available. e.g. the code will put the log (conditionally) gave 10% discount, thus derived value of 90 (100 - 10) reducing $10 of the calculated value above = 80 (90 - 10) Basically, having a log of the sequence of steps, the state on which the code relies, input any function takes and output thereafter (but in a user understandable manner or in business language). One step closer to this could be writing code in a language that is relevant to business I feel we still write our code in terms of technology (arrays, exceptions, serializations etc). I don’t mean that we should not worry about those things. But, those things should be covered under a layer which makes easy for business people OR a new comer to come back to the code and read it like a specification.
Nagarjun 13 Dec 2009 10:50 am: We faced same situation while trying to find out how the comission is calculated. In the end it so happened that the commision was being calculated based on data that was 5 years old. Nobody had realised that the new data was not comin into the system :)

The Long Tail of information sharing

Thu, 27 Apr 2006 12:00:00 +0000

The Long Tail of information sharing. Even on Wikipedia, fewer than 2% have contributed over 100 articles. Over 85% have contributed fewer than 5 articles. A Wiki inside an organisation is unlikely to reach critical mass, left to itself.

Comments

Arun 28 Apr 2006 6:10 am: But I imagine, many in the 85% probably contributed stuff that the prolific guys never thought of. For instance, I know one guy who put up something on Ananth Pai (Tinkle, Amar Chitra Katha), and i doubt he probably ever added anything else. But you are right about wiki in an organization. Do you know an organization where a wiki has been used effectively?
S Anand 28 Apr 2006 10:22 pm: No, not offhand. We were planning to recommend it at one of our clients’, but I’m not going to do that any more.
sathish 30 Apr 2006 2:39 am: if the organization strength is large - runs into many thousands or lakhs - I think the required critical mass might be acheived.. for smaller companies and small teams too, it might make sense - since, every one is liable to participate - especially product documentation - it would be easier make/correct and change.
Arun 30 Apr 2006 6:02 am: Ah, ok.
S Anand 30 Apr 2006 8:45 am: A colleague just reported that his current client is successfully using a Wiki. So may it’s not impossible after all…
Madhu 2 May 2006 12:24 pm: wiki is all about scale i think. The sergery brin special lecture at UCB is a case in point. You can access this at google videos

Information Age helps the forgetful

Fri, 28 Feb 2003 12:00:00 +0000

Information Age Intelligence talks about how “… the information age … opens up the opportunity for those with weaker memories to compete on a more even playing field than those with good ones.” And it’s the one of the best things that ever happened to me. via andersja

McKinsey blocks competitors

Wed, 13 Nov 2002 12:00:00 +0000

An interesting thread on Andersja’s about how McKinsey is blocking competitors from its site. I’m at the receiving end of it. And so far, it hasn’t affected my life in any way.

The Social Life of Documents

Tue, 01 Oct 2002 12:00:00 +0000

The Social Life of Documents. Interesting article by Xerox’s PARC on the role of documents in the New Economy.