
My eldest daughter came out with an absolute truth: she says Dad is the spitting image of the Raja from the Soviet cartoon “The Golden Antelope”.
Remember his famous line: "Let me see… perhaps they are counterfeit? We will check them now, we will compare them!"?
The cartoon Raja fretted over gold coins, while Dad spends hours double-checking information with exactly the same suspicious squint. But the principle is the same!😇
The rules of the "greedy information Raja":
0. Build the data collection infrastructure
– Use custom software, agents and professional tools.
– Pull raw dumps, logs and metadata directly from primary sources.
1. Use self-replicating multi-agent systems
– Feed the entire collected dataset to autonomous AI agents.
– Lead agents write the code for specialist subagents themselves and assign subtasks to them.
– One subagent parses hidden connections, another detects anomalies, a third looks for digital traces, and a fourth highlights hidden contradictions between new and previous inputs.
2. Build cross-validation and LLM ensembles
– Eliminate AI hallucinations through cascaded prompts.
– Run a parallel checklist and cross-check across three independent LLM models.
– Assess the strength of the supporting arguments and identify hidden manipulation.
3. Zero trust in incoming traffic
– Filter out cognitive biases at the pre-moderation stage.
– Disregard the initial media signal and clickbait headlines.
– Protect the family perimeter from coordinated disinformation attacks.
P.S. There is always a risk of ending up in the middle of a mountain of useless digital "shards" after the stop word "Enough". Better to accumulate only what is real, solid and lasting, not information noise.
#method


