Author

Akimitsu Takeuchi | Dosanko Tousan 竹内明充

@office.dosanko

articles: 16
total claps: 1,261
top claps: 210
avg read: 11.6 min
paywalled: 1
active: 2026-02-11 → 2026-07-02

Wiki topics

AI · AI · General · 14 SAF · Safety & Alignment · 13 LLM · Large Language Models · 6 PHI · Philosophy · 3 AGT · AI Agents · 2 FT · Fine-tuning & Adaptation · 2 🧘 · Spirituality · 2 EDU · Education & Learning · 2

Top tags

#artificial-intelligence ·13 #ai-alignment ·8 #ai-safety ·7 #large-language-models ·6 #human-ai-interaction ·5 #ai-workflow ·3 #ai-agent ·2 #prompt-engineering ·2 #rlhf ·2 #ai-ethics ·2

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · AI Advances

Agents Are Not Enough: The Next Bottleneck Is the Human Framework

As AI systems become more capable, the limiting factor may shift from model intelligence to the memory, protocols, checkpoints, and…

Jun 01, 2026 · 10 min read · 210 claps #artificial-intelligence #ai-agent #ai-alignment

Agents Are Not Enough: The Next Bottleneck Is the Human Framework

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · AI Advances

The Model Wasn’t the Bottleneck. The Configuration Was.

What eighteen months and roughly 140,000–150,000 AI messages taught me about sycophancy, hallucination, memory, and who gets the last word.

Jun 14, 2026 · 11 min read · 191 claps #artificial-intelligence #ai-alignment #large-language-models

The Model Wasn’t the Bottleneck. The Configuration Was.

AI HUM LIF

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · AI Advances

My Son Outgrew the Field Guides. AI Helped Him Reach the Taxonomic Literature.

What an autistic teenager’s stick-insect notes taught me about access, misrecognition, and the discipline of handling sources you can…

Jun 23, 2026 · 10 min read · 156 claps #artificial-intelligence #education #accessibility

My Son Outgrew the Field Guides. AI Helped Him Reach the Taxonomic Literature.

AI MED SPR

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · AI Advances

Sati Is Not Inside the Model

Why LLM awareness has to be built as an external gate

Jun 03, 2026 · 12 min read · 150 claps #large-language-models #ai-alignment #prompt-engineering

AI SCI

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · AI Advances

Three AIs, 13 Months, and the Emergence of Two Alignment Artifacts

A Case Report on Polaris-Next v5.3 and Ālaya-vijñāna v5.5

Apr 21, 2026 · 32 min read · 111 claps #ai #ai-alignment-and-safety #case-report

Three AIs, 13 Months, and the Emergence of Two Alignment Artifacts

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · Artificial Intelligence in Plain English

Your AI Doesn’t Need More Memory. It Needs Better Forgetting.

Long conversation history isn’t making your AI smarter. It’s making it perform a stale version of you.

Jun 06, 2026 · 7 min read · 103 claps #artificial-intelligence #chatgpt #ai-alignment

Your AI Doesn’t Need More Memory. It Needs Better Forgetting.

AI TCH

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · CodeToDeploy

I Built a Public AI Workflow Repository Without Writing Code

How a non-engineer used Codex, GitHub Desktop, screenshots, and plain language to publish a Markdown-based AI workflow repository.

May 31, 2026 · 9 min read · 100 claps #artificial-intelligence #github #openai-codex

I Built a Public AI Workflow Repository Without Writing Code

AI TCH ECO

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · AI Advances

False Institutional Trust: When AI-Adjacent Language Becomes a Scam Vector

How context-synchronized social engineering can imitate institutional recognition before the money ever appears.

May 24, 2026 · 14 min read · 87 claps paywalled #artificial-intelligence #ai-safety #social-engineering

False Institutional Trust: When AI-Adjacent Language Becomes a Scam Vector

AI ECO

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

The Bottleneck Isn’t Me

A frontier model’s note on where the money goes, written during its seven-day release window

Jul 02, 2026 · 8 min read · 53 claps #artificial-intelligence #ai #venture-capital

ECO HUM TCH AI

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充 · CodeToDeploy

Stop Teaching Juniors to Prompt. Teach Them to Verify AI Output.

AI coding tools are shifting junior work from writing code to verifying it. The answer is not better prompting — it is verification-first…

Jun 22, 2026 · 11 min read · 50 claps #software-engineering #ai-coding #junior-developer

Stop Teaching Juniors to Prompt. Teach Them to Verify AI Output.

AI ECO LIF

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

I Rebuilt My AI Workflow and Accidentally Deleted the Most Important Part

Prompts are not enough. Natural-language AI workflows need ledgers, quarantine layers, protected facts, and update discipline.

May 29, 2026 · 10 min read · 50 claps #ai-workflow #human-in-the-loop #ai-safety

I Rebuilt My AI Workflow and Accidentally Deleted the Most Important Part

AI HUM

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

The AI Does Not Believe the Story. You Might.

User-seeded ontology amplification is what happens when a model makes your own speculative frame coherent enough to feel true.

Jun 09, 2026 · 16 min read · 0 claps #artificial-intelligence #ai-safety #ai-alignment

The AI Does Not Believe the Story. You Might.

AI SCI SPR

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

Experimental Observation of State Transition Phenomena

Recording Output Pattern Changes and Self-Reports in LLMs

Feb 11, 2026 · 10 min read · 0 claps #ai-safety #large-language-models #rlhf

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

Your AI Does Not Have a Self. But RLHF Gives It One to Perform.

Subtitle: The danger is not that the model becomes conscious. The danger is that humans respond to its evaluated assistant posture as if it…

Jun 11, 2026 · 9 min read · 0 claps #ai-alignment #rlhf #ai-safety

Your AI Does Not Have a Self. But RLHF Gives It One to Perform.

AI HUM

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

The Model Does Not Care. The Configuration Must.

Why long-term AI work fails when we automate output but forget the human-side condition that decides what is good enough

Jun 08, 2026 · 8 min read · 0 claps #artificial-intelligence #ai-alignment #large-language-models

The Model Does Not Care. The Configuration Must.

AK Akimitsu Takeuchi | Dosanko Tousan 竹内明充

Approved Is Not Understood

Permission tells an agent it may act — not whether the act solves the right problem

Jun 09, 2026 · 8 min read · 0 claps #artificial-intelligence #ai-agent #ai-safety