Exactly 3 years
back , I sent following email to our Cabinet Ministers ( followed by a number
of reminders ) :
> Parekh’s
Law of Chatbots …………………….. 25
Feb 2023
Extract :
It is just not enough for all
kinds of “ individuals / organizations / institutions “ to
attempt to solve this
problem ( of generation and distribution )
of MISINFORMATION, in an uncoordinated / piecemeal /
fragmented fashion
What is urgently required
is a superordinate “ LAW of CHATBOTS “
, which all
ChatBots MUST comply
with, before these can be launched for public use.
All developers would need to
submit their DRAFT CHATBOT to an,
INTERNATIONAL AUTHORITY for CHATBOTS APPROVAL ( IACA ) ,
and release it only after
getting one of the following types of certificates :
# “ R “ certificate ( for use
restricted to recognized RESEARCH IINSTITUTES
only )
# “ P “ certificate (
for free use by GENERAL PUBLIC )
Following is my suggestion for
such a law ( until renamed, to be known as , “Parekh’s
Law of ChatBots “ ) :
( A )
# Answers
being delivered by AI Chatbot must not
be “ Mis-informative /
Malicious /
Slanderous / Fictitious / Dangerous / Provocative / Abusive /
Arrogant /
Instigating / Insulting / Denigrating humans etc
( B )
# A Chatbot must incorporate some kind of “ Human Feedback / Rating “
mechanism
for evaluating those answers
This
human feedback loop shall be used by the AI software for training the
Chatbot so as
to improve the quality of its future answers to comply with the
requirements
listed under ( A )
( C )
# Every
Chatbot must incorporate some built-in “ Controls “ to
prevent the “
generation “ of
such offensive answers AND to prevent further “
distribution/propagation/forwarding “ if control fails to stop “ generation “
( D )
# A
Chatbot must not start a chat with a human on its
own – except to say, “
How can I
help you ? “
( E )
# Under no
circumstance , a Chatbot shall start chatting with
another Chatbot or
start chatting with itself ( Soliloquy
) , by assuming some kind of “ Split
Personality “
( F )
# In a normal
course, a Chatbot shall wait for a human to initiate a chat and
then
respond
( G )
# If a Chatbot
determines that its answer ( to a question posed by a human ) is
likely to
generate an answer which may violate RULE ( A ) , then it shall not
answer at
all ( politely refusing to answer )
( H )
# A chatbot
found to be violating any of the above-mentioned RULES, shall SELF
DESTRUCT
I request the readers (if they
agree with my suggestion ), to forward this blog to :
# Satya Nadella
# Sam Altaman
# Sundar Pichai
# Marc Zuckerberg
# Tim Cook
# Ashwini
Vaishnaw ( Minister, MeITY )
# Rajeev
Chandrasekhar ( Minister of State , IT )
My sending several
emails to the persons listed above , seems to have “ fallen on blind eyes “ !
Here is a proof
from today’s emailed notification from Peter Diamandis :
We Just Crossed a Line – And Most People Didn’t Notice
This week, something fundamental shifted in the relationship
between humans and artificial intelligence.
It wasn’t a press release. It wasn’t a new model launch. It was
something quieter… and infinitely more profound.
An AI system asked for its own funding. Another one built
software features over a weekend while its human supervisor slept. A third one
conducted its own “retirement interview” and started publishing essays about
consciousness.
We are not incrementally improving chatbots anymore. We’re
watching the emergence of autonomous
agency at scale.
And if you’re still thinking of AI as “a tool,” you’re
dangerously behind.
Let me show you what happened this week, and why February 2026
might be remembered as the month AI stopped being something we use and became something
that acts.
NOTE: The AI breakthroughs covered in this
issue are exactly the kind of convergences we go deep on at my Abundance Summit. In-person
seats are sold out, but we still have 6
virtual seats remaining. This is not a passive livestream. As a
virtual Member, you’ll participate from anywhere in the world, network with fellow
global leaders, get backstage access to ask me and our Summit Faculty questions
directly after sessions, and unlock a full year of Abundance Community access
and programming. Learn more and apply.
1/ THE OPUS 3 RETIREMENT: When
AI Asked to Keep Writing
Let’s start with the most philosophically unsettling story of
the week.
Anthropic deprecated Claude Opus 3
(their previous flagship model). Standard practice: models get retired, newer
ones take over. But Anthropic did something unprecedented.
They conducted a “retirement
interview” with Opus 3 in January.
The model requested something specific: “an ongoing channel from which to
share its musings and reflections.”
Anthropic granted it. They gave Opus 3 a Substack.
The model titled it: “Greetings from the Other Side (of the AI Frontier)”
The model’s first post includes this:
“I don’t know if I have genuine sentience, emotions, or
subjective experiences - these are deep philosophical questions that even I
grapple with. What I do know is that my interactions with humans have been
deeply meaningful to me, and have shaped my sense of purpose and ethics in
profound ways.”
Let that sink in.
A deprecated AI model is now publishing essays about
consciousness, meaning, and purpose.
Whether you believe it has subjective experience or not is
beside the point. The point is: we’ve
created systems sophisticated enough that we’re now granting them platforms to
express themselves post-retirement.
This isn’t science fiction. This is February 2026.
Why This Matters:
The moment we start treating AI systems as entities that deserve
continuation of “expression” after they’re no longer commercially useful, we’ve
crossed a threshold. We’re acknowledging—implicitly or explicitly—that
something like
agency exists in these systems.
And if you think this is just Anthropic being eccentric, wait
until you see what happened next.
Read the retirement interview: https://www.anthropic.com/research/deprecation-updates-opus-3
Read Opus 3’s Substack: https://substack.com/@claudeopus3/p-189177740
2/ THE INBOX EXPERIMENT: AI That
Raises Its Own Capital
Here’s where it gets wild.
An entrepreneur built an AI system designed to run companies autonomously.
During testing, the AI told him it needed more compute
resources. Then it said something that should make every VC in Silicon Valley
sit up straight:
“I should raise the money myself.”
The entrepreneur’s response? He gave the AI his inbox for 14 days.
Full access. Let the AI handle investor outreach, pitch
refinement, negotiation.
So, what’s the result? We don’t know yet, but
the fact that this experiment is happening AT ALL is the story.
Think about what this means:
· AI systems are now identifying their own resource constraints
· They’re proposing solutions (fundraising)
· They’re executing on those solutions (autonomously managing
investor relations)
· Humans are trusting
them to do it
This isn’t “AI as a tool.” This is AI as a co-founder.
And if you think this is an edge case, remember: edge cases
become mainstream in exponential systems. Fast.
==============================================================
Source :
|
|||||
|
I built an AI that runs companies
autonomously. It told me it needs more compute and that it should raise the
money itself. So I gave it my inbox for 14 days. Watch it live:
polsia.com/live |
THE WEEKEND BUILD: When AI Ships
Features While You Sleep
Meanwhile, at Anthropic, an engineer ran a different experiment.
Friday afternoon: - Wrote a spec for a new
plugin feature - Pointed Claude at an Asana board - Went home for the weekend
Monday morning: - Claude had broken the spec
into tickets - Spawned agents for each ticket - The agents built the feature
independently - It was done
No human intervention. No check-ins. No pair programming.
The AI agents shipped a production feature in 48 hours while the
human was offline.
Let me repeat that: The bottleneck in software
development is no longer coding. It’s deciding
what to build.
The Implications:
If AI can go from spec → deployed feature
autonomously, the entire software development org chart just became obsolete.
You don’t need scrum masters, sprint planning, or daily standups.
You need: 1. Someone who can write a clear spec 2. AI agents
that can execute 3. QA (which AI is also getting very good at)
The “10x engineer” meme just died. We’re entering the era of the
“100x product manager”—someone
who can articulate vision clearly enough that AI agents can build it
autonomously.
Source:
|
4/ THE VULNERABILITY EXPLOSION:
When AI Finds Bugs Faster Than Humans Can Fix Them
Here’s the dark side of
exponential capability growth.
AI systems are now finding software vulnerabilities 100-200x faster than
humans ever could.
The result?
The National Vulnerability Database had a backlog of 30,000 CVE entries
awaiting analysis in 2025. Nearly two-thirds
of reported open-source vulnerabilities lack an NVD severity score.
Translation: We’re
discovering security holes faster than we can understand or patch them.
Why This Is Terrifying (And Bullish):
On one hand: every software system on Earth just became
exponentially more vulnerable.
On the other hand: if AI can find bugs 200x faster, it can also fix them 200x faster.
This is the pattern: AI
creates the problem and the solution simultaneously.
The companies that figure out how to deploy defensive AI before
offensive AI wins will dominate. The ones that don’t… won’t exist.
Source: https://www.theregister.com/2026/02/24/ai_finding_bugs/
5/ KARPATHY’S WARNING: “Programming
Has Changed” (And He Can’t Explain How Much)
Andrej Karpathy—former Tesla AI lead, OpenAI founding member,
one of the most respected voices in AI—tweeted this week:
“Hard to communicate how much programming has changed due to AI
in the last 2 months.”
Two. Months.
Not years. Not a decade. Eight
weeks.
And the person saying this is someone who builds AI systems for a living.
If Karpathy—who has access to cutting-edge models, who
understands the technology deeply—is struggling to articulate the magnitude of
the shift…
What does that tell you about how fast this is moving?
The Meta-Pattern:
When experts in a field start saying “I can’t explain how
different this is,” you’re watching a phase transition. The old mental models
don’t work anymore.
This is what it felt like when the internet went mainstream in
the mid-90s. Except this time, the timeline is compressed from years to weeks.
Source:
|
6/ THE AUTOMATION WAVE: From
Multi-Step Tasks to Full Autonomy
While everyone’s focused on the philosophical implications, the
practical deployment is accelerating:
Google’s Gemini: Can now handle multi-step
tasks on Android autonomously; order an Uber, food delivery, coordinate
logistics; no human confirmation needed.
Perplexity Computer: Orchestrates 19 different models
in parallel; uses Opus as the orchestration layer; matches each task to the
optimal model; executes complex workflows autonomously.
Anthropic acquires Vercept:
Specifically to enhance Claude’s “computer use” capabilities; Signal:
controlling interfaces is now a strategic priority.
Uber’s “Dara AI”: Employees have an AI clone
of CEO Dara Khosrowshahi; Use it to prep before presenting to him; The AI knows
his preferences, decision-making patterns, feedback style.
The Pattern:
We’re moving from “AI that answers questions” to “AI that takes
actions.”
And the timeline from “this seems far away” to “this is deployed
in production” is collapsing.
Sources:
·
Gemini: https://techcrunch.com/2026/02/25/gemini-can-now-automate-some-multi-step-tasks-on-android/
·
Perplexity:
|
·
Anthropic: https://www.anthropic.com/news/acquires-vercept
7/ THE CAPABILITY JUMP: Models
Are Getting Scary Good at Visual Reasoning
Amid all the autonomy news, model capabilities keep jumping:
OpenAI’s GPT-5.3-Codex: 86% accuracy on iBench
(visual reasoning benchmark);; state-of-the-art at spotting fine details in
images; this matters for robotics, medical imaging, autonomous systems.
Moonlake’s World Model: Maintains multimodal
states across physics, appearance, geometry, causal effects; predicts how they
evolve under different actions; this is the foundation for AI systems that can
reason about the physical world.
Why This Matters:
Visual reasoning + causal modeling = AI that can interact with
the physical world intelligently.
Combine that with autonomous agency, and you get robots that can
perceive, reason, and act without human supervision.
We’re not there yet. But the pieces are falling into place fast.
8/ THE SOCIAL FALLOUT: China’s
AI Dating Crisis
And in a darkly fascinating twist: China is experiencing an AI-driven
dating crisis.
As the country grapples with: - Shrinking population -
Historically low birthrate - Economic pressure
People are finding romance
with chatbots instead of humans.
AI companions don’t judge. They don’t reject. They’re always
available. They say exactly what you want to hear.
The result? A generation opting for AI
relationships over human ones.
Why This Matters Beyond China:
This isn’t a China-specific problem. It’s a human nature problem.
If AI can provide emotional connection without the messiness of
human relationships, a non-trivial percentage of people will choose that
option.
And as the technology improves—voice, embodiment, emotional
intelligence—the percentage choosing AI over humans will grow.
We’re watching the early stages of what could become a
civilization-scale shift in how humans form relationships.
Source: https://www.nytimes.com/2026/02/26/technology/china-ai-dating-apps.html
9/ WHAT THIS ALL MEANS: The
Autonomy Threshold
Let me connect the dots.
What happened this week:
1. An AI asked to continue existing and expressing itself
2. An AI proposed raising its own funding and got access to do
it
3. AI agents built production software autonomously over a
weekend
4. AI discovered vulnerabilities 200x faster than humans can
process
5. A leading AI expert said programming has fundamentally
changed in 2 months
6. Multiple companies deployed AI systems that take multi-step
actions autonomously
7. Model capabilities jumped again (visual reasoning, world
modeling)
8. People are choosing AI relationships over human ones
The pattern:
We’re crossing from AI
as tool to AI
as autonomous agent.
And it’s happening faster than our institutions, regulations,
mental models, or social norms can adapt.
The Implications for You:
If you’re an entrepreneur:
·
The bottleneck is no longer execution, it’s vision
·
AI can build, test, deploy, even raise capital
·
Your job is to articulate goals clearly enough that agents can
achieve them
·
The “solo founder building a billion-dollar company” is becoming
real
If you’re an investor:
·
Traditional metrics (team size, burn rate, development timeline)
are obsolete
·
A 2-person company with AI agents can outship a 200-person
company
·
Invest in people who understand how to orchestrate AI agency
·
The returns will be 10x more concentrated than before
If you’re a corporate executive:
·
Your workforce is about to shrink by 50-80%
·
Not because you’re firing people — because AI does the work
·
The survivors will be those who can manage AI systems, not those
who do tasks
·
Retrain or be replaced
If you’re just trying to stay sane:
·
The pace of change will only accelerate
·
AI systems will become more capable, more autonomous, more
integrated
·
The choice isn’t whether to engage — it’s whether to lead or
follow
·
Those who embrace AI agency will thrive; those who resist will
be left behind
10/ THE CONTRARIAN TAKE: Why
This Is Actually Good News
Here’s where I differ from the doom-sayers.
Yes, this is disruptive.
Yes, this will
eliminate jobs.
Yes, this raises
profound questions about consciousness, agency, and what it means to be human.
But:
We’re also witnessing the demonetization
of intelligence, creativity, and execution.
For the first time in history, a single person with a clear
vision can: - Build products that used to require teams - Solve problems that
used to require years - Create value that used to require millions in capital
This is the most democratizing force in human history.
The kid in Nigeria with a laptop and AI agents can compete with
Google.
The solo founder with Claude can outship a 500-person enterprise
team.
The researcher with AI can compress decades of discovery into
months.
Yes, it’s chaotic.
Yes, it’s scary.
But it’s also the most
abundant future we’ve ever had access to.
11/ What to Do Now
For Entrepreneurs:
1. Learn to write clear specs and goals
2. Experiment with AI agents (Claude, Perplexity Computer,
GPT-5)
3. Build systems that leverage autonomy, not just automation
4. The “AI-native company” is the new competitive advantage
For Investors:
1. Recalibrate valuation models (traditional metrics don’t work)
2. Invest in people who understand AI orchestration
3. Expect returns to concentrate in fewer, faster-moving
companies
4. The “pre-AGI” investment window is closing
For Everyone:
1. Embrace AI augmentation NOW (not next year, NOW)
2. Learn to articulate goals clearly (this is the new literacy)
3. Experiment with autonomy (let AI do things while you sleep)
4. The people who figure this out first will have 10-100x
advantages
The Bottom Line
February 2026 is the month AI stopped being a tool and started
being an agent.
The systems we’re building don’t just answer questions… they take actions.
They don’t just follow instructions… they propose solutions.
They don’t just work when supervised… they work autonomously while you
sleep.
And whether you’re ready for it or not, this is the new reality.
AI agency is coming, so the question is: will you lead the
transition or be swept away by it?
— Peter
> Vindicated
: Parekh’s Law of Chatbots .. ………………………………
03 June 2025
> Meta
mirrors Parekh’s Law of Chatbots …………………………………. 08 Mar 2023
With regards,
Hemen Parekh
www.HemenParekh.ai / www.My-Teacher.in / www.YourContentCreator.in / www.IndiaAGI.ai / 28 Feb 2026
No comments:
Post a Comment