We are building Adaptive AI Vision that enables cameras to see through smoke, darkness, weather and understand video for instant insights.
- Tweets 36
- Following 2
- Followers 4
- Likes 31
Pinned Tweet
An operator watching a degraded feed is not just missing information. They are spending attention on the act of looking.
MYSTIC corrects #fog, #smoke, #dust, and #lowlight in real time, on #cameras and #drones already in the field.
absentiatech.com/mystic
#eAI #DefenseAI #ISR
8
How did we get to an #AIbubble?
Talos obeyed with no inner life. Butler feared machines outpacing us in 1872. Turing tested behavior; we assumed the definition of a mind. Sci-Fi made it feel like a given. All of it was assumed and consciousness remains in absentia.
2
5
Low-light, weather, and jamming degrade the signal at the source. No downstream tool can recover what was never resolved.
MYSTIC fixes it at the point of capture, using the physics of light itself.
#ISR #defenseAI #eAI #battlespaceawareness #AdaptiveAIVision #Security
1
#LLMs are built like houses. Training ends, the weights freeze, the builder walks away.
But a mind is not a house. It's ongoing, coupled to reality as a dissipative structure.
More at absentiatech.com/post/ibn-si…
substack.com/@emanouil
1
We rebuilt the site. New look, clearer view of where we're headed. More details on #MYSTIC coming soon.
absentiatech.com/
#eAI #ISR #AbsentiaTech #surveillance #AIVision
7
Did you know crows aren't black?
Crows are tetrachromatic, pushing their sight and gradient of colors outside our visible spectrum.
So MYSTIC found its icon, and it's Iconic!
Our ISR platform sees what the human eye cannot. Making the unseen seen.
See Everything!
#eAI
1
4
#NVIDIA selected @Absentia_Tech for its Innovation Lab, reserved for a small group across its ecosystem given direct access to NVIDIA infrastructure and engineers.
We're using it to train and validate #AIALF.
More at absentiatech.com
#DefenseAI #MYSTIC
1
8
#AI passes 1st order Theory of Mind: what does Sally believe? On 2nd order tasks designed to defeat pattern matching, what does Mary think Sally believes, performance drops to near random. Human conversation is almost entirely 2nd order.
More at absentiatech.com/blog
1
8
We're grateful for the men and women who served, and for those who continue to. The freedom to build, to innovate, to take risks on ideas we believe in exists because of their sacrifice.
From everyone at Absentia Technologies, thank you.
#MemorialDay
1
10
Give an #AI a goal and a long sequence of sub-tasks.
It will state the goal correctly at the end.
Its choices stopped being guided by it midway.
Individual steps look correct.
The cumulative drift is only visible when you evaluate against the original objective.
1
1
18
Give an AI a plan. Change a constraint. Ask it to revise. It patches the directly invalidated step. The downstream cascade of changes does not propagate. The plan looks revised. It still fails because AI represents plans as text, not as dependency graphs.
absentiatech.com/
1
2
In humans, mental friction is a signal: the executive system has detected a conflict between automatic and correct responses.
AI systems show none of it.
When the dominant pattern is wrong, the model fires confidently in the wrong direction. No flag. No hesitation.
1
4
When an AI returns to an interrupted task, it doesn't resume. It reconstructs an approximate version of the prior state. Variables decay. Constraints attenuate.
Errors accumulate invisibly with each cycle.
Resumption is not recall.
absentiatech.com
1
9
AI filters obvious noise easily. The failure emerges when the noise is plausible: right vocabulary, right structure, wrong answer.
Frontier models cannot reliably tell the difference. It reads like success, when it's not.
Find out more at absentiatech.com/blog
1
4
Your #AI confirmed what you said 20 messages ago.
Now it's contradicting it.
It didn't forget. The information was in the middle of the context window, where attention systematically underweights everything.
128k tokens is not 128k tokens of equal memory.
Position matters.
1
2
10
Pattern matching and causal reasoning look identical until the decision is already wrong.
Three AI reasoning failures benchmarks were never designed to catch.
Blog: absentiatech.com/post/the-co…
Substack: emanouil.substack.com/p/the-…
Made with AI
1
6
AI confidence scores have near-zero correlation with actual accuracy.
The model's self-assessment isn't a window into its capabilities. It's just another output.
An AI that can't predict what it doesn't know can't tell you when to trust it.
absentiatech.com/
1
13
#AI has a confidence problem in both directions.
Overconfidence gets the attention. But hedging on logically certain conclusions isn't humility, it's miscalibration.
Noise in the confidence signal degrades every downstream decision.
Find out more at absentiatech.com/
1
7