https://aicassindra.com/blogs/adk/adk_agent_types.htmlhttps://i.ytimg.com/vi/awKfXINTFIA/hqdefault.jpgChatbot vs AI Agent: one question shows the difference (Short)one question, a chatbot and an agent side by sidehttps://www.youtube.com/embed/awKfXINTFIA2026-10-21T12:20:48-07:00https://i.ytimg.com/vi/ErBOjEbdWBk/hqdefault.jpgWorkflow vs AI agent: who decides the next step? (Short)a workflow fixes the order; an agent picks the next stephttps://www.youtube.com/embed/ErBOjEbdWBk2026-10-23T11:33:26-07:00 https://aicassindra.com/blogs/adk/adk_circuit_breaker.htmlhttps://i.ytimg.com/vi/klkhxcJm7rc/hqdefault.jpgStop building AI agents without limits (Short)cap model calls before a loop runs awayhttps://www.youtube.com/embed/klkhxcJm7rc2026-10-21T12:20:45-07:00 https://aicassindra.com/blogs/adk/adk_events.htmlhttps://i.ytimg.com/vi/AqMQi5Lsm7Y/hqdefault.jpgOne prompt vs 11 steps: inside an AI agent (Google ADK) (Short)the events one prompt really produceshttps://www.youtube.com/embed/AqMQi5Lsm7Y2026-10-21T12:20:42-07:00 https://aicassindra.com/blogs/adk/adk_guardrails.htmlhttps://i.ytimg.com/vi/qhOJ3TeVDEc/hqdefault.jpgModel vs your code: who guards an AI agent's tool? (Short)the check belongs in code, not in the modelhttps://www.youtube.com/embed/qhOJ3TeVDEc2026-10-29T06:30:24-07:00 https://aicassindra.com/blogs/adk/adk_guardrails_pattern.htmlhttps://i.ytimg.com/vi/MyXqIG4fQ-Q/hqdefault.jpgPrompt vs code: where AI agent guardrails belong (Short)a prompt asks; code enforceshttps://www.youtube.com/embed/MyXqIG4fQ-Q2026-10-23T04:32:20-07:00 https://aicassindra.com/blogs/adk/adk_llm_agent.htmlhttps://i.ytimg.com/vi/awKfXINTFIA/hqdefault.jpgChatbot vs AI Agent: one question shows the difference (Short)one question, a chatbot and an agent side by sidehttps://www.youtube.com/embed/awKfXINTFIA2026-10-21T12:20:48-07:00 https://aicassindra.com/blogs/adk/adk_overview.htmlhttps://i.ytimg.com/vi/mbW22lSlu4g/hqdefault.jpgGoogle ADK Java Tutorial: Build Your First AI Agent (Ep 1)What an AI agent really is, and how ADK runs onehttps://www.youtube.com/embed/mbW22lSlu4g2026-10-21T11:27:22-07:00 https://aicassindra.com/blogs/adk/adk_rag_grounding.htmlhttps://i.ytimg.com/vi/R1Ydpt2BJew/hqdefault.jpgStop your AI agent from making things up (Short)ground answers in data, not guesseshttps://www.youtube.com/embed/R1Ydpt2BJew2026-10-21T12:20:38-07:00 https://aicassindra.com/blogs/adk/adk_runtime.htmlhttps://i.ytimg.com/vi/AqMQi5Lsm7Y/hqdefault.jpgOne prompt vs 11 steps: inside an AI agent (Google ADK) (Short)the events one prompt really produceshttps://www.youtube.com/embed/AqMQi5Lsm7Y2026-10-21T12:20:42-07:00https://i.ytimg.com/vi/klkhxcJm7rc/hqdefault.jpgStop building AI agents without limits (Short)cap model calls before a loop runs awayhttps://www.youtube.com/embed/klkhxcJm7rc2026-10-21T12:20:45-07:00 https://aicassindra.com/blogs/adk/adk_tool_design_patterns.htmlhttps://i.ytimg.com/vi/ErW7xz9bQRE/hqdefault.jpgStop trusting your AI agent's tool arguments (Short)validate what the model passes to a toolhttps://www.youtube.com/embed/ErW7xz9bQRE2026-10-29T07:34:54-07:00https://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/adk/adk_tools.htmlhttps://i.ytimg.com/vi/6LqHoQ5lCOA/hqdefault.jpgAI agent tool vs Java method: same thing (Google ADK) (Short)a tool is a plain Java methodhttps://www.youtube.com/embed/6LqHoQ5lCOA2026-10-22T11:37:15-07:00https://i.ytimg.com/vi/ErW7xz9bQRE/hqdefault.jpgStop trusting your AI agent's tool arguments (Short)validate what the model passes to a toolhttps://www.youtube.com/embed/ErW7xz9bQRE2026-10-29T07:34:54-07:00https://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/adk/adk_workflow_agents.htmlhttps://i.ytimg.com/vi/ErBOjEbdWBk/hqdefault.jpgWorkflow vs AI agent: who decides the next step? (Short)a workflow fixes the order; an agent picks the next stephttps://www.youtube.com/embed/ErBOjEbdWBk2026-10-23T11:33:26-07:00 https://aicassindra.com/blogs/adk_java/adk_java_agent_unit_testing.htmlhttps://i.ytimg.com/vi/8LS1ztXwx84/hqdefault.jpgReal model vs fake model: how to test an AI agent (Short)test the agent with a scripted fake modelhttps://www.youtube.com/embed/8LS1ztXwx842026-10-23T11:34:44-07:00 https://aicassindra.com/blogs/adk_java/adk_java_callbacks.htmlhttps://i.ytimg.com/vi/MyXqIG4fQ-Q/hqdefault.jpgPrompt vs code: where AI agent guardrails belong (Short)a prompt asks; code enforceshttps://www.youtube.com/embed/MyXqIG4fQ-Q2026-10-23T04:32:20-07:00 https://aicassindra.com/blogs/adk_java/adk_java_debugging_hallucinations.htmlhttps://i.ytimg.com/vi/R1Ydpt2BJew/hqdefault.jpgStop your AI agent from making things up (Short)ground answers in data, not guesseshttps://www.youtube.com/embed/R1Ydpt2BJew2026-10-21T12:20:38-07:00 https://aicassindra.com/blogs/adk_java/adk_java_testing_custom_llms.htmlhttps://i.ytimg.com/vi/8LS1ztXwx84/hqdefault.jpgReal model vs fake model: how to test an AI agent (Short)test the agent with a scripted fake modelhttps://www.youtube.com/embed/8LS1ztXwx842026-10-23T11:34:44-07:00 https://aicassindra.com/blogs/adk_java/adk_java_token_counting_across_models.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/blogs/adk_java/adk_java_tool_error_wrapping.htmlhttps://i.ytimg.com/vi/qhOJ3TeVDEc/hqdefault.jpgModel vs your code: who guards an AI agent's tool? (Short)the check belongs in code, not in the modelhttps://www.youtube.com/embed/qhOJ3TeVDEc2026-10-29T06:30:24-07:00https://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/adk_java/adk_java_tools_advanced.htmlhttps://i.ytimg.com/vi/6LqHoQ5lCOA/hqdefault.jpgAI agent tool vs Java method: same thing (Google ADK) (Short)a tool is a plain Java methodhttps://www.youtube.com/embed/6LqHoQ5lCOA2026-10-22T11:37:15-07:00https://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/agents/agents_tool_use.htmlhttps://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/agents/tool_calling_best_practices.htmlhttps://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/llm/llm_structured_output.htmlhttps://i.ytimg.com/vi/uLL6R4nwITk/hqdefault.jpgGoogle ADK Function Tools in Java: Validate Tool Arguments (Ep 2)Function tools with explicit contractshttps://www.youtube.com/embed/uLL6R4nwITk2026-10-28T10:23:47-07:00 https://aicassindra.com/blogs/llm/llm_tokenization.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/blogs/llm/llm_tokenizer.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/blogs/llm/understanding_tokenization.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/blogs/transformer_math/tm_bpe_deep.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/blogs/transformers/tokenization_choices.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/blogs/transformers/transformers_tokenizers.htmlhttps://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/labs/chat_template_renderer.htmlhttps://i.ytimg.com/vi/2KEbP_OGzFk/hqdefault.jpgQwen vs Gemma chat templates: one refuses your system prompt (Short)the same conversation through two chat templateshttps://www.youtube.com/embed/2KEbP_OGzFk2026-11-02T19:00:00+05:30https://i.ytimg.com/vi/HPYAUCQLhBE/hqdefault.jpgKannada Was Always Ready. AI Wasn't: 14x → 1.8x TokensTokens: what you are really paying forhttps://www.youtube.com/embed/HPYAUCQLhBE2026-10-30T22:30:24-07:00 https://aicassindra.com/labs/llm/kv_cache_memory.htmlhttps://i.ytimg.com/vi/L1qMBzIg9c8/hqdefault.jpgReserved vs paged KV cache: 18 vs 246 requests on the same 2 GiB (Short)how many requests fit in 2 GiB of KV cachehttps://www.youtube.com/embed/L1qMBzIg9c82026-10-31T00:30:38-07:00 https://aicassindra.com/labs/llm/observability_adk_java_eval_framework_overview.htmlhttps://i.ytimg.com/vi/mBADKEBPsuo/hqdefault.jpgAggregate score vs per-slice gate: 85% hid a 50% (Short)an 85% average hiding a 50% slicehttps://www.youtube.com/embed/mBADKEBPsuo2026-11-01T07:30:16-08:00 https://aicassindra.com/labs/llm/optimization_adk_guardrails_pattern.htmlhttps://i.ytimg.com/vi/k4Op3O1vT-w/hqdefault.jpgStop guarding only the prompt: 1 of 4 harms stopped (Short)a prompt check alone against four harmshttps://www.youtube.com/embed/k4Op3O1vT-w2026-11-03T19:00:00+05:30 https://aicassindra.com/labs/llm/speculative_decoding.htmlhttps://i.ytimg.com/vi/sbKcJ0-0FhU/hqdefault.jpgSpeculative decoding vs plain: 2.6x faster, then 0.78x slower (Short)when drafting tokens pays off and when it does nothttps://www.youtube.com/embed/sbKcJ0-0FhU2026-11-03T13:00:00+05:30 https://aicassindra.com/labs/llm/temperature_sampling.htmlhttps://i.ytimg.com/vi/4MgCieVFjVY/hqdefault.jpgTemperature 0.2 vs 2.0: Paris every time, then Lyon (Short)60 draws at two temperatures, run in this labhttps://www.youtube.com/embed/4MgCieVFjVY2026-10-31T08:30:08-07:00 https://aicassindra.com/labs/llm/top_k_top_p.htmlhttps://i.ytimg.com/vi/00HS4p_SYJ0/hqdefault.jpgTop-k vs top-p: 8 wrong tokens vs 0 in 600 draws (Short)600 draws with top-k and with top-phttps://www.youtube.com/embed/00HS4p_SYJ02026-11-02T13:00:00+05:30 https://aicassindra.com/labs/prompt_injection_scanner.htmlhttps://i.ytimg.com/vi/mELegfe4PM0/hqdefault.jpgEnglish vs Kannada: the AI guardrail only spoke English (Short)the same attack, caught in English and missed in Kannadahttps://www.youtube.com/embed/mELegfe4PM02026-10-30T08:30:38-07:00 https://aicassindra.com/labs/rag_chunker.htmlhttps://i.ytimg.com/vi/WXwUvAt0388/hqdefault.jpgCharacter vs sentence chunks: RAG split 3 of 4 answers (Short)fixed-size chunks cut 3 of 4 answers in twohttps://www.youtube.com/embed/WXwUvAt03882026-11-01T00:30:10-07:00