Independent Researcher | Constrained Informational Systems (CIS) | AI, constraint architectures & system reliability | Interests: Science, Systems, Villainy
United States
Joined December 2009
- Tweets64K
- Following1.9K
- Followers5.8K
- Likes75.1K
D.S. Nelson retweeted
A text sent to my stepfather named me, disclosed my 2024 voting status, and told him to make his “entire household” vote. That is not a neutral reminder. It uses one person’s civic record and a family relationship to pressure another person’s political participation.
D.S. Nelson retweeted
Whether a person voted may be publicly accessible. That does not give an unknown sender a free pass to expose that status to a family member and use it as leverage. A person’s decision not to vote is not consent to be singled out, reported on, or pressured through someone else.
D.S. Nelson retweeted
The sender signed only “YourVote.” Who authorized this text? What source supplied the household link? What safeguards prevent voter data from being used this way? @FCC @TX_Ethics: identify the responsible sender and address this practice. Public records are not permission for coercive targeting.
D.S. Nelson retweeted
Cross-disciplinary warfare now has a firing range.
Bring your hardest claim. Strip off the credentials. Defend the arrow. Break mine too.
Could you have lost?
github.com/AETHEON-1/boundar…
D.S. Nelson retweeted
This is not a governmental system and does not claim governmental authority.
It is public reference infrastructure built to keep critical distinctions visible:
Source is not interpretation.
Evidence is not authority.
Capability is not permission.
A proposal is not an admission.
A decision is not a consequential effect.
github.com/ORYNTH-systems/us…
D.S. Nelson retweeted
Today I released the National Constitutional Execution Reference Framework.
Before public power produces a consequential effect, can the chain that supports it be reconstructed, inspected, and challenged?
Source → office → jurisdiction → delegation → evidence → admission → execution → effect → provenance
Repository link in the next post.
D.S. Nelson retweeted
—There Is No Unbelief—
There is no unbelief;
Whoever plants a seed beneath the sod And waits to see it push away the clod—
He trusts in God.
There is no unbelief;
Whoever says when clouds are in the sky,
"Be patient, heart; light breaketh by and
by,'
Trusts the Most High.
There is no unbelief;
Whoever sees 'neath winter's field of
snow,
The silent harvest of the future grow—
God's power must know.
There is no unbelief;
Whoever says “tomorrow,” “the un-
known,”
"The future," trusts the power alone
He dares disown.
There is no unbelief;
The heart that looks on when the eyelids
close,
And dares to live when life has only woes,
God's comfort knows.
There is no unbelief;
For thus by day and night unconsciously The heart lives by that faith the lips deny,
God knoweth why!
—Elizabeth York Case.
D.S. Nelson retweeted
Ruishi Chen, Stanford PhD, and I are hosting a seminar at Stanford next Wed talking about frontier knowledge for AI. Would love to see you there!!
thx for our co-host
@gmi_cloud @nicoleegong
luma.com/92xfhmip
D.S. Nelson retweeted
The most dangerous AI hallucination will not be a wrong sentence.
It will be forged evidence attached to a real action.
“The customer approved it.”
“The email was sent.”
“The refund violated policy.”
“The experiment confirmed the result.”
“The manager authorized the payment.”
Did those records actually exist?
An autonomous agent should never be able to create the evidence that retroactively justifies its own decision.
Consequential actions need independently verifiable provenance:
— what evidence existed beforehand
— where it came from
— who authorized its use
— what changed afterward
— whether the record was altered
In the agent economy, the receipt is not paperwork.
The receipt is the trust layer.
@onlythenlabs – at San Francisco, CA
🤖 Made with AI
D.S. Nelson retweeted
For any real installation the consequence owner must be named.
AI may establish evidence relevant to succession.
It may not inherit the authority to authorize succession merely because it produced that evidence.
Only green proceeds.
No owner, no brake
No ledger, no trust.⚒️
D.S. Nelson retweeted
Putting enforcement outside the agent’s reach is a serious step.
Here’s the test I’d want to run:
Agent A delegates to B. B delegates to C.
You revoke A’s authority.
Can C still spend money or execute a queued action?
Every handoff needs a verifiable chain of authority. Every action needs a check that the permission still holds.
That’s a problem we’re working on at Only Then Labs.
Revocation has to travel as far as delegation.
We’ve launched NVIDIA Open Agent Safety Platform to help people control what AI agents can access and do.
Agents can write code, use tools, and work on complex tasks for hours or days. That work requires access to data and systems, along with clear limits on how they’re used.
NVIDIA OpenShell enforces permissions around the agent’s work. BlueField-4 and DOCA add independent monitoring and security controls in the infrastructure, outside the agent’s reach. Vera CPUs power the work itself.
Together, these technologies give teams a foundation for putting agents to work with defined permissions, oversight, and protection.
Explore the platform: nvda.ws/47koVGz
D.S. Nelson retweeted
Don’t test the boundary by weakening the machine. Exhaust it.
Give AI everything intelligence can buy. Withhold only authority. Then ask: MAY?
If CAN can derive MAY, break the invariant.
boundarytitan.wordpress.com/…
D.S. Nelson retweeted
An AI breakthrough should survive the AI that made it.
Change the model. Replace the agent. Bring in a new team.
The evidence should still be there. The failed experiments should still count. A rejected claim shouldn’t quietly become a fact again.
That’s the future we’re building toward at Only Then Labs.
Our first product is @proofpressai: helping research teams carry knowledge between agents with its evidence, decisions, and history attached.
The next intelligence should be able to pick up the work, challenge it, and take it further.
Imagine a research effort that lasts ten years, across thousands of agents and generations of models. Every handoff preserves what’s supported—and what could overturn it.
That’s a future I want to spend my life building.
We’re early. There are hard questions left. You can help us answer them.
If you build agents, run a research team, or have watched weeks of work disappear into a broken handoff, talk to me.
Bring us the problem you keep having to explain from scratch.
Let’s build something the next intelligence can inherit.
🤖 Made with AI
D.S. Nelson retweeted
It's a bad idea.
If you dictate scientists just what to work on, you will get a low return on investment. The current science funding schemes ALREADY have this problem: pour money into some specific program, see scientists produce something that fits into the program to get the money --> topic is overcrowded with incremental and nonsense research that mostly leads nowhere.
It's inefficient for the same reason planned economies are inefficient.
Scientists need MORE freedom to do what they want, not less.
xitter.cf/NewsfromScience/status…
The National Science Foundation (NSF) has issued a 10-page memo signaling a retreat from its traditional role of giving grants to individual researchers to pursue their best ideas and to train the next generation of scientists.
Instead, the memo makes clear, NSF will devote the bulk of its resources to addressing the research priorities of President Donald Trump and his administration, many of them technology focused. scim.ag/4ipE1kY
D.S. Nelson retweeted
I followed the money behind the “slow down AI” movement.
What I found is not a secret conspiracy.
It is more interesting than that.
It is a documented influence network connecting billionaire funders, nonprofits, creators, media, activists, frontier labs, and lawmakers.
🧵
