google on S Anand

Fri, 18 Apr 2025 13:10:15 +0000

With the Gemini 2.5 Flash release, Google envelopes the entire cost-quality frontier of LLMs. In other words, at any cost or quality level, today, the best model to use according to the LM Arena score is a Gemini model.

Results for O3, O4 Mini, and GPT 4.1 are not yet on LM Arena. But until then, #Google dominates. Nice work!

Link: https://sanand0.github.io/llmpricing/

Fri, 20 Sep 2024 04:07:16 +0000

Looks like XML tags are the best way to structure prompts and separate sections for an #LLM. It’s the only format that all of Anthropic, Google, and OpenAI LLMs encourage.

For example:

… … … …

Anthropic Docs: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/use-xml-tags OpenAI Docs: https://platform.openai.com/docs/guides/prompt-engineering/strategy-write-clear-instructions Google Docs: https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/structure-prompts

Alternatives are using JSON, Markdown, templating formats like Mustache/Jinja, etc.

Even Llama’s system tokens seem a little XML-like. https://github.com/meta-llama/llama3/blob/main/llama/tokenizer.py#L61-L74

Personally, I’ve been using Markdown so far. But it’s time to switch over. (Only on the prompt side. On the generation side, Markdown still seems the best.)

Short notes

Wed, 20 May 2009 05:15:32 +0000

I’m quite busy on a project right now, and don’t get time to write long articles. So for a while, I’m going to stick to short notes on interesting stuff.

Peter Bregman has a very interesting piece on Why You Should Encourage Weakness. It boils down to a choice: do you focus on on improving strengths or minimising weaknesses? Conventional performance evaluations focus on the latter. I very strongly support Bregman’s view on this. The weakness isn’t why you hired the person! Unless it’s killing the organisation, just leave them to focus on their strengths.
Google Analytics has a fairly interesting API that I hadn’t explored until recently. Picked up [Advanced Web Metrics with Google Analytics](http://www.s-anand.net/amazon-browser.html#advanced web metrics with google analytics) and learnt that you can track outbound clicks, page load times, Javascript events and error logs, almost anything at all using Google Analytics. You can also mirror the logging on your local server using pageTracker._setLocalRemoteServerMode()
The whole concept of a Sandbox environment seems to be picking up within Google. There’s a Checkout sandbox, an AJAX API playground, an AdWords sandbox, an AdSense API sandbox, the Mapstraction API sandbox, even an event called Developer Sandbox. (After saying Sandbox 6 times, I feel a bit like Hobbes.)

Attack of the bots

Sun, 31 Aug 2008 12:00:00 +0000

One out of every 5 hits to my site is from a bot.

I spent a fair bit of time this weekend analysing my log file for last month (which runs to gigabytes, and I ended up learning a few things about file system optimisation, but more on that later). 80% of the hits were from regular browsers. 20% were from robots. Here's a sample of the user-agents:

Mozilla/5.0 (compatible; Yahoo! Slurp; <a href="http://help.yahoo.com/help/us/ysearch/slurp)">http://help.yahoo.com/help/us/ysearch/slurp)</a>
Mozilla/5.0 (compatible; Googlebot/2.1; +<a href="http://www.google.com/bot.html)">http://www.google.com/bot.html)</a>
Mediapartners-Google
DotBot/1.0.1 (<a href="http://www.dotnetdotcom.org/#info">http://www.dotnetdotcom.org/#info</a>, crawler@dotnetdotcom.org)
Mozilla/5.0 (Twiceler-0.9 <a href="http://www.cuill.com/twiceler/robot.html)">http://www.cuill.com/twiceler/robot.html)</a>
msnbot/1.1 (+<a href="http://search.msn.com/msnbot.htm)">http://search.msn.com/msnbot.htm)</a>
FeedBurner/1.0 (<a href="http://www.FeedBurner.com)">http://www.FeedBurner.com)</a>
Mozilla/5.0 (compatible; attributor/1.13.2 +<a href="http://www.attributor.com)">http://www.attributor.com)</a>
WebAlta Crawler/2.0 (<a href="http://www.webalta.net/ru/about_webmaster.html)">http://www.webalta.net/ru/about_webmaster.html)</a> (Windows; U; Windows NT 5.1; ru-RU)
Yandex/1.01.001 (compatible; Win16; I)
...

You get the idea. The bulk of these are search engines. Over two-thirds of the bot requests were from Yahoo Slurp. Now, this struck me as weird. If I take the top 3 search engines that are sending traffic my way,

	Referral %	Crawl %
Google	90%	24%
Yahoo	6%	66%
Microsoft	3%	0.3%
Others	1%	9%

The search engine that sends me the most traffic is being reasonably conservative, while Yahoo is just eating up the bandwidth on my site. Actually, this shouldn't bother me too much. It's not taking up too much bandwidth, or even CPU usage, given that all the bots put together make up only 20% of my traffic. But somehow... it's sub-optimal. Inelegant, even.

So I decided to take a closer look. Just how often are they crawling my site?

Yahoo	*Every 5 seconds*
Google	Every 13 seconds
DotBot	Every 9 minutes
Cuill	Every 9 minutes
Microsoft	Every 18 minutes
Feedburner	Every 18 minutes
Attributor	Every 23 minutes
Yandex	Every 27 minutes

Look at those numbers. Yahoo is hitting my site once every 5 seconds. No wonder there's a help page at Yahoo titled How can I reduce the number of requests you make on my web site? I followed their advice and set the crawl-delay to 60, so at least it slows down to once a minute.

Just that one little line change should (hopefully) reduce the load on my site by around 15%.

As for the other engines, I don't mind that much in terms of load.

Google, for all that it crawls every 13 seconds, has faithfully reported that it has only 11% of my site under its index, so I've no idea what they're doing, but I'm not complaining about the traffic that's coming my way.
DotBot. Today was the first I'd heard of them. Visited the site, and smiled. These guys can do all the crawling of my site that they like, and I hope something interesting comes out of their work.
Cuill, sends me 0.2% of my traffic, but it's a new search engine, I'm happy to give it time.
Microsoft's OK, sends me a tiny stream of traffic.
Feedburner is just pinging my RSS feed every 18 minutes.
Attributor and Yandex I'm hearing of for the first time, again. Not too much load on a system, so that's OK.

What's amazing is the sheer number of bots out there. Last month, I counted over 600 distinct user-agent strings just representing bots. So it's true. The Web is no longer just for humans. We do need a Semantic Web.

Comments

Dhar 31 Aug 2008 9:09 pm: Hmmm, curious as to:
1. Why the bots should crawl your site every 5 seconds or so?
2. How you can find out how much of your site has been indexed by Google.
Cheers, D.
S Anand 1 Sep 2008 12:11 am:
1. I think Yahoo’s crawler is aggressive in any case. My site doesn’t seem to be an exception: there are a lot of threads discussing this problem.
2. Google’s webmaster tools tells you how many URLs have been indexed from your sitemap.

The Next Big Language

Sun, 11 Feb 2007 12:00:00 +0000

Steve Yegge at Google talks about the features of the Next Big Language. He apparantly has inside information about the language corporates are likely to make a big push for. The comments seem to suggest Javascript 2.

Timeline of Microsoft Google and Yahoo acquisitions

Mon, 13 Nov 2006 12:00:00 +0000

A timeline of Microsoft, Google and Yahoo acquisitions.

London Test Automation Conference at Google

Sun, 24 Sep 2006 12:00:00 +0000

Presentations and videos from Google’s conference on test automation at London. I was at the event, and learned a lot.

Google internal subdomains

Wed, 20 Sep 2006 12:00:00 +0000

Google’s internal subdomains.

Google Master Plan

Tue, 22 Aug 2006 12:00:00 +0000

Google’s master plan.

Videos you can learn from

Thu, 20 Jul 2006 12:00:00 +0000

Berkeley webcasts of their courses.
Google TechTalks.
Authors@Google.
LongNow seminars about long term thinking.
UCTV Video on Demand.
Nova.
Computer History Museum.

Comments

Michelle 5 Dec 2006 12:39 pm: Do you have books I can learn From? Top 10 non-fiction books?

How Google was named Google

Fri, 07 Jul 2006 12:00:00 +0000

5 different stories about how Google was named Google. This is mentioned as the probably correct version:

“Lucas Pereira: ‘You idiots, you spelled [Googol] wrong!’ But this was good, because google.com was available and googol.com was not. Now most people spell ‘Googol’ ‘Google’, so it worked out OK in the end.”

Comments

OverSkies 11 Dec 2008 8:29 am: Google’s founder’s son’s first words were : google.That is why they named it google

Google meta search

Tue, 04 Jul 2006 12:00:00 +0000

What do you find when you search Google searches?

Comments

Sheikh 14 Dec 2006 3:09 pm: Thank You Anand. It was very useful

55 ways to have fun with Google

Fri, 23 Jun 2006 12:00:00 +0000

55 ways to have fun with Google.

Comments

Dhar 25 Jun 2006 7:59 am: Just finished the book, didn’t find it too interesting. :((
S Anand 25 Jun 2006 8:52 am: Me neither. :-(

Sketchup is free

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

Sketchup (from Google) is now available for free download. It’s a 3D modelling tool.

Determine a sex by name

Mon, 17 Apr 2006 12:00:00 +0000

Determine a sex by name. If you don’t know if a name is male or female, just search for the name on Google images. (e.g. “Priti”)

Comments

Umasuthan 18 Apr 2006 5:36 pm: That is not always true. Try ‘Kiran’
S Anand 18 Apr 2006 10:17 pm: You’re right. The article does warn of unisex names!

Google web authoring statistics

Wed, 25 Jan 2006 12:00:00 +0000

Google web authoring statistics. An analysis of over a billion pages to see how people use HTML markup.

Who is afraid of Google

Sat, 03 Dec 2005 12:00:00 +0000

Who’s afraid of Google? Everyone

Blind Search Engine Test

Mon, 28 Nov 2005 12:00:00 +0000

The Blind Search Engine Test. My vote turned out to be for Google.

Google may acquire Riya

Thu, 17 Nov 2005 12:00:00 +0000

Google may acquire Riya. Riya can recognise faces and lettering in your pictures.

Comments

Sai 8 Dec 2005 11:50 pm: Hi, do post the rest of the tips for excel when you get the time. Good Stuff.

Talent Wars

Mon, 31 Oct 2005 12:00:00 +0000

Talent wars. The interesting part was the first three paragraphs.

Flying on the Delta Shuttle with Bill Gates 12 years ago, I asked, “What Microsoft competitor worries you most?”

“Goldman Sachs.” I gave Gates a startled look. Was Microsoft about to try the investment banking business? “Software,” he said, “is an IQ business. Microsoft must win the IQ war, or we won’t have a future. I don’t worry about Lotus or IBM, because the smartest guys would rather come to work for Microsoft. Our competitors for IQ are investment banks such as Goldman Sachs and Morgan Stanley.”

I spent five days traveling the country with Gates, and he must have talked about IQ a hundred times. Getting the brightest bulbs to work at Microsoft has always been his obsession. It’s paid off. Microsoft does close to $40 billion in sales and has some 60,000 employees. That’s a whopping $650,000-plus of revenue per employee, topping IBM’s sales per employee twofold.

Along comes Google, with its revenue run rate of $6 billion and about 4,000 employees. Google’s sales per employee are $1.5 million, or 2.3 times that of Microsoft. This is like comparing Babe Ruth to Home Run Baker. Google now beats Microsoft in the IQ war.

Comments

Jayant 3 Nov 2005 2:46 am: Just an opinion. I think there is subtle difference between the two talents. Both are sharp but one is more commercially driven. I do think lot of behaviour changes happen due to that orientation