Title: P1: TOON (Tuple-Oriented Object Notation) [2026-01-10 Sat]

"name": "Team Alpha",
"members": [
{ "id": 1, "name": "Alice" },
{ "id": 2, "name": "Bob" }
]
}
]
}
```

```ton
teams[1]:
- name: Team Alpha
members[2]{id,name}:
1,Alice
2,Bob
```

Links:
- specification http s:// github. com/toon-format/spec/blob/main/SPEC.md
- article http s:// openapi. com/blog/what-the-toon-format-is-token-oriented-object-notation
⚰ #dailyreport #tupleformat #llmformat #json

Title: P0: TOON (Tuple-Oriented Object Notation) [2026-01-10 Sat]

Indentation based format.
"TOON or TON" is good for tiny nesting. For flat CSV format, for big nesging JSON is the best.

Terms:
- Header - termintad with “:”
- Field list: Brace-enclosed, delimiter-separated list of field names for tabular arrays: {f1f2}.
- List item: A line beginning with "- " at a given depth representing an element in an expanded array.

```json
{
"teams": [
{ #dailyreport #tupleformat #llmformat #json

Title: P1: Offtopic, dont hurt me. what is ... time? [2026-01-09 Fri]

of time from quantum entanglement between a system and
an ancillary quantum clock. Smith and Ahmadi (2019)
extended this to interacting systems, noting that
physical systems typically interact, with gravity
coupling to all matter.
https://www.nature.com/articles/s41586-024-08005-8
⚰ #dailyreport #time #spacetime #physics #fundamentals

Title: P0: Offtopic, dont hurt me. what is ... time? [2026-01-09 Fri]

Genius IT rock-star is looking for this.
Why it is irreversable, sharp and why it actually move.
Waves are irreversable also.
For sharp the answer that it is a https://en.wikipedia.org/wiki/Relativity_of_simultaneity and not sharp.
But for “why it move” the answer is (following I don't understand)
"The Page and Wootters mechanism” describes the emergence #dailyreport #time #spacetime #physics #fundamentals

Title: P1: Releasing my app for chat with LLM [2026-01-09 Fri]

environment. There is no hidden formatting, everything
is open and visible, giving you full control. This is
how professionals work with LLMs. We use open source
and control everything. It's not a mobile-like Mac or
whatever toy you use.

OAI supports several providers, including Gemini and
OpenAI.
⚰ #dailyreport #emacs #llm #aiassitant #ai #llmchat #openai

Title: P0: Releasing my app for chat with LLM [2026-01-09 Fri]

I released the first stable version of my package, OAI,
for the GNU Emacs text editor on the MELPA platform.

At left part of screenshot is the main window with chat
and AI answer. At right part is a help window with
expanded link.

This package provides a chat window for communication
with LLMs or any AI agent, all within a pure text #dailyreport #emacs #llm #aiassitant #ai #llmchat #openai

Title: P3: Own Thoughs about LLM certainty and Uncertainty [2025-12-25 Thu]

- Symbolic AI: invented reasoning paths when rules don’t cover the case.
- Bayesian Models: inflated posteriors built on weak priors.
- Expert Systems: template projection onto unknown inputs.
- Reinforcement Learning: reward illusions under noisy signals.
- Decision Trees: fabricated boundaries when data is thin.

https:/ /www.linkedin. com/in/gavin-tam-35621923/
≡ω≡ #hallucination #conswerative #llm #aiengineering

Title: P2: Own Thoughs about LLM certainty and Uncertainty [2025-12-25 Thu]

- input clarity for model

I suggest:
- Model Certainty = ModelComplexity + fitting + TrainingCertainty
- Output Certainty = ModelCertainty - (penalization - InputUncertainty)

Penalization or guirdrails is as a part of input.

In post there was interesting list of "Uncertainty for
ML":

- Evolutionary AI: premature convergence when the search space is incomplete. #hallucination #conswerative #llm #aiengineering

Title: P1: Own Thoughs about LLM certainty and Uncertainty [2025-12-25 Thu]

If the system knows less → Δ grows.

Delta Δ means residual Error (true - output) of congruence.

I think, "Hallucination" as result of
- Overfitting - simple answers for hard questions
- Underfitting - inadequate reasoning for simple questions.

To measure Hallucination we should count:
- model complexity
- penalization of uncertainty - Overfitting/Underfitting #hallucination #conswerative #llm #aiengineering

Title: P0: Own Thoughs about LLM certainty and Uncertainty [2025-12-25 Thu]

I found not bad post at Linkedin, that states new
"The Certainty Conservation Law" for systems.
Total_Output_Certainty = Input_Information_Certainty
+ Manufactured_Certainty_Δ

Hallucination as 'Entropy Compensation'.

This is same was as fundamental Conservation Law: "there
is something that not change in system over time."

Δ ≈ Information_Gap × Output_Pressure #hallucination #conswerative #llm #aiengineering

Title: P3: Infinity, duality, time [2025-12-23 Tue]

about how ideas evolve.
≡ω≡ #infinity #math #duality #dao #laozi #genesis #time

Title: P2: Infinity, duality, time [2025-12-23 Tue]

There is no god because nothing can get out of the trap
of space-time layers.

In the Bible, Genesis says that "heaven and earth"
(surface) were first created, and then the surface was
isolated. How are dimensions based on infinity and a
dot?

Lao Tzu and Genesis say that time is cycled as light and
dark. What does this mean?

AI says: time isn’t just about when things happen — it’s #infinity #math #duality #dao #laozi #genesis #time

Title: P1: Infinity, duality, time [2025-12-23 Tue]

Laozi says — but they are so fundamentally different,
like different dimensions.

Time occurs at all levels and all distances at once,
undivisible.

Zeno paradox solution:
- Potential - we halve the distance.
- Actual - all the halvings already existing at once.
So in reality there is no iteration or tasks or
sequence?, only relations. #infinity #math #duality #dao #laozi #genesis #time

Title: P0: Infinity, duality, time [2025-12-23 Tue]

I think the Bible's Genesis ✞ is a commentary on Laozi’s
*Tao Te Ching* ☯.

The main duality of reality: the current moment and
distance-time, like a dot and infinity. No one has seen
them or imagined them, but if reality is real, then
infinity and the edge are real.

The convergence of infinity to some real value shows us
how infinity and the dot "go from one to another" as #infinity #math #duality #dao #laozi #genesis #time

Title: P1: Main prompt engineering rule or what LLM remembers. [2025-12-17 Wed]

See the attached screenshot with a conversation from my
"OAI" environment for chat in the GNU Emacs
editor. https:/ /github. com/Anoncheg1/emacs-oai

LLM refers to chat history as "earlier prompts", "items
from our conversation", "our current session",
"previous messages" or "previously asked".
≡ω≡ #dailyreport #prompt #promptengineering #llm #aispeak #aicontext #llmcontext #emacs #contextengineering

Title: P0: Main prompt engineering rule or what LLM remembers. [2025-12-17 Wed]

LLM does not remember text in context and does not
recognize it if it was not marked in any way.

If you don't ask LLM to remember something, or if it
does not carry some common markers/meanings for the
chat, then LLM will not recognize it and will not use
it afterwards. #dailyreport #prompt #promptengineering #llm #aispeak #aicontext #llmcontext #emacs #contextengineering

Title: P2: DeepCode - code generation or doc-to-repos. [2025-12-12 Fri]

link https:/ /arxiv. org/pdf/2512.07921

Another paper was reviewed by me at [2025-08-22]:
2024 𝗣𝗮𝗽𝗲𝗿𝟮𝗖𝗼𝗱𝗲: 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗰 𝗖𝗼𝗱𝗲 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 𝗳𝗿𝗼𝗺
𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗣𝗮𝗽𝗲𝗿
≡ω≡ #dailyreport #codegeneration #doctorepos #coding #codingassistant

Title: P1: DeepCode - code generation or doc-to-repos. [2025-12-12 Fri]

3) Verification protocol: setup, metrics, success
criteria
4) Execution environment: dependencies
5) Stages of development: order and staged integrated
verification
- Code generator *CodeMem* — builds memory as entries:
purpose, interface, internal and external dependencies
- Code generator *CodeRAG*
- Testing #dailyreport #codegeneration #doctorepos #coding #codingassistant

Title: P0: DeepCode - code generation or doc-to-repos. [2025-12-12 Fri]

In this paper, emphasis is placed on controlling context
and attention for agents empowered with RAG.
- Concept agent
- Algorithm agent
- Planning - ensuring that every abstract component is
directly linked to a precise technical
specification. Outputs:
1) File hierarchy
2) Component specs: modules, classes, functions with
mathematical formulation #dailyreport #codegeneration #doctorepos #coding #codingassistant

Title: P1: P2 Reading LLM As a Judge, subagents [2025-12-12 Fri]

- Better monitoring
- Improved tool selection for multi-step tasks
- Easy to test components in isolation
Cons:
- The system can be rigid
- The orchestrator must manage each step
≡ω≡ #dailyreport #llmasajudge #subagents #agenticai #aiagent #reasoning