I wrote and sold my first computer games age 13 in Assembler on the C64 late 80ies.
Today I am - on the technical side - DBA, #Database-guy and #Golang Fan today.
I built the SaaS software LinkResearchTools, and Link Detox, and most recent, URLinspector.
A software developer, turned marketer, turned software developer - etc. - some circles :-).
I also run the Mastodon instances
developer.social
digitalmarketing.social
c64.chat
Some "experts" say
"You don't need prompts for AI
because
we now have plugins to access databases."
That's like saying:
We don't need programming languages,
because we have our data stored in databases.
Just a confused conclusion,
equalling two completely separate things
out of incompetence:
instructions and data.
It's not either or.
I love writing me a fancy regex when editing complex stuff.
But why bother writing regex...
...when you can get the same text transformation with a good GPT-4 prompt?
#writing #transformation #editing #ai #prompts #programming #coding #softwareengineering
Founder skills nobody in the "Maker community" talks about, but everyone needs
1⃣ Trademark law
2⃣ Copyright law
3⃣ "Cost of doing business" attititude
(lawyers, credit card fees, risk management, backup X, protective legal actions, domain registrations, etc)
4⃣ Email Sending
We now live in a world where you can casually write new multi-core performance benchmark tools with GPT-4 while waiting a couple minutes for your ansibles to re-image and re-install your new cluster hardware.
The new hardware setup will be good for 6 to 10 million users then, if all the benefits compound as expected...
Now, with 1 million users in AIPRM, we'll soon hit the estimated limit of 1.5 million of the old test hardware we run AIPRM on.
We will give AIPRM its own cluster, no more sharing resources with LRT 😉
I just confirmed deployment details of the new AIPRM cluster, using hardware not available to the public yet.
* Double Node Count in Cluster
* Latest CPU Top Model
* Latest Memory DDR5-ECC replacing DDR4-ECC
I am soooo excited...
A "database" that only has a client library for Python and NodeJS?
Are you kidding me?
OpenAI had 7500 nodes in Kubernetes in 2021,
BEFORE GPT-3 got really popular,
and of course
BEFORE the "Iphone Moment" of ChatGPT
in November 2022 and
BEFORE the huge $10B investment by Microsoft paid on Azure credits mostly.
How many nodes & GPUs do they run today?
Not disclosed, for good reasons mostly.
But just training GPT-5 takes supposedly 25,000 GPU nodes.
https://openai.com/research/scaling-kubernetes-to-7500-nodes
Just remembered that a client once asked in a Site Clinic, if it would make sense to
setup his Wordpress (PHP Scripts on MySQL)
...in a Kubernetes Cluster.
LOL, no way.
Some "recommendation" that was.
Moving some PHP scripts to a K8S cluster won't make them better, faster or anything.
Virtualizing that is like installing Linux to run Windows.
It just results in some web dev being able to add "K8s" to his resume, without 100% lying.
Just 99.9% 🤣
👉AIPRM Prompt Template Language for VS Code
A prompt is not 1 sentence. 🤓
Prompts are the new code. 🥸
And so here is the VS Code Language extension
for the
👉AIPRM Prompt Template Language
An important preparation for even more powerful and free features coming soon... ;-)
I've been benchmarking 3 new server models, and it is amazing.
FYI, AIPRM still runs our ca 2 yr old reference cluster hardware and does it very smoothly with 800k users.
We will be fine going to 2 million users, at ease, with the current QA cluster we use.
But just replacing the bare metals with the most recent models, we can go to 10M users on the same node count.
Piece of cake with good IT architecture and binaries instead of interpreted scripts on the server.
So glad we got our Google OAuth limit lifted.
We almost hit 50k Google OAuth token grants per day yesterday.
Thanks to everyone who helped with Retweeting, you rock!
🥰😍🤩😘🤗
We need someone at Google to help us to connect more than 10,000 new users/day with their API, we still have default Google OAuth limit.
Do YOU know anyone that could help?
The metering that AIPRM applies to both features and quotas is a very generous one.
The prices set, starting from $2 are considered early-bird prices and may be increased for those buying after April 17.
No we are not making money selling ads or user data, like Google or Facebook or LinkedIn or Twitter, and so just charge for our work, our intellectual property, our server resources, our legal cost and so much more.
Nothing wrong about that.
Welcome to all #AIPRM Premium users!
For some, your (free) "mind-blowing solution" becomes "gimmicky" once it costs $1, even if that's just for additional features on top of fascinating free value still provided.
For some, your time, effort, resources, and personal attention are "supposed" to be free. (?)
Luckily, those 0.000001% are just outliers.
We're just getting started with AIPRM! 🚀 😎 🤓
More links & comments in the AIPRM Community Forum
https://forum.aiprm.com/t/google-new-ai-ecosystem-announcement-vertex-ai/3580
Huge Google AI Announcement on Pi Day
- PaLM API & MakerSuite
- AI in Gmail, Google Docs & workspace
- Generative AI support in Vertex AI
- Generative AI App Builder
- plus: new PalM-E that got burried on Friday due to SVB
Wow, that’s a lot of Google AI announcement in 1 day, plus partnerships, programs, and resources for each segment of the ecosystem.
OpenAI has disabled the ChatGPT history over claimed security issue.
This has nothing to do with AIPRM, but it also happens without any extensions
AIPRM has no power to fix this issue,
it is up to OpenAI
More background and updates:
2 days after closing the AIPRM Facebook group, we have more visits to the 48-hour young forum,
on our owned content and platform,
than we got referals from Facebook in
8 weeks 😉😏☺️
Just keep feeding walled gardens, folks.
We are SEOs and in for the traffic and exposure.