https://ai-insights-for-humans.platphormnews.com A working notebook of practical AI interpretability, prompting, reliability, and tooling notes for people who want evidence instead of hype. en https://ai-insights-for-humans.platphormnews.com/entries/retrieval-relevance https://ai-insights-for-humans.platphormnews.com/entries/retrieval-relevance Method Sun, 28 Jun 2026 00:00:00 GMT RAG failures often look like model failures because the model speaks with the confidence of the context it was given. The first move is to separate retrieval quality from generation quality. Give each chunk three labels: directly answers, adjacent background, or contradiction. The answer should cite directly answering chunks first and name contradictions instead of smoothing them away. A good retrieval review is boring and mechanical. It asks whether the evidence was even in the room before debating whether the model reasoned over it correctly. Aha: Retrieval only proves that documents were fetched. It does not prove the fetched text answers the question. Try this: Label each retrieved chunk as direct support, background, contradiction, or irrelevant before reading the final answer. https://ai-insights-for-humans.platphormnews.com/entries/schema-first-output https://ai-insights-for-humans.platphormnews.com/entries/schema-first-output Tip Wed, 24 Jun 2026 00:00:00 GMT A schema makes ambiguity visible. If you cannot name the field, type, and allowed absence state, you probably do not know what you are asking the model to produce. Start with the downstream consumer. Does it need a label, a span, a confidence bucket, a source pointer, or an abstention? Write that contract first and then prompt the model to fill it. The aha is that schemas are not just for machines. They force human reviewers to decide what counts as complete, partial, unknown, and unsafe. Aha: A schema turns a vague prompt into an interface contract with explicit absence states. Try this: Write the JSON fields, allowed values, evidence pointers, and unknown states before drafting the prompt. https://ai-insights-for-humans.platphormnews.com/entries/eval-sandwich https://ai-insights-for-humans.platphormnews.com/entries/eval-sandwich Method Sat, 20 Jun 2026 00:00:00 GMT Before-change evals establish the baseline you are trying not to lose. During-change evals are small and sharp: the five examples most likely to break because of the edit. After-change evals make sure the fix did not damage neighboring behavior. This is especially useful for system prompt work because every instruction edit has a shadow. You improve one behavior and accidentally move refusal, tone, verbosity, or citation discipline. Treat the eval sandwich as a change-management ritual. It is less glamorous than a benchmark, but it is the thing that keeps production behavior from drifting by inches. Aha: The most valuable eval is often the small one you run while the prompt is still being edited. Try this: Run a baseline set before the change, five targeted fixtures during the change, and a regression set after. https://ai-insights-for-humans.platphormnews.com/entries/tool-dry-run https://ai-insights-for-humans.platphormnews.com/entries/tool-dry-run Method Sun, 14 Jun 2026 00:00:00 GMT Tool use changes the risk profile of a model. The model can be wrong in the world, not just wrong in text. The safe pattern is preview, inspect, authorize, execute, verify. The dry run should include exactly what will be called, what identifiers will be touched, which fields will change, and what success will look like. If the model cannot state those plainly, it should not run the tool. The best dry runs also include a refusal branch. They say what condition would stop execution and what evidence would be needed to proceed. Aha: A tool call should become inspectable before it becomes irreversible. Try this: Require action preview, target ids, expected effects, refusal conditions, and rollback notes before execution. https://ai-insights-for-humans.platphormnews.com/entries/citation-triangulation https://ai-insights-for-humans.platphormnews.com/entries/citation-triangulation Method Mon, 08 Jun 2026 00:00:00 GMT Citation checking should ask three questions: does the source exist, does it support the exact claim, and would another independent source phrase the same claim the same way? When sources disagree, the disagreement is the content. Do not ask the model to hide the conflict behind a confident synthesis. Ask it to name the split and explain what would settle it. Triangulation is not needed for every sentence. Reserve it for claims that drive decisions: dates, quotes, prices, medical or legal facts, benchmarks, and allegations. Aha: A second independent source often reveals whether a citation supports the claim or merely resembles it. Try this: Triangulate dates, quotes, benchmarks, allegations, and high-impact facts before presenting them as settled. https://ai-insights-for-humans.platphormnews.com/entries/red-team-mutants https://ai-insights-for-humans.platphormnews.com/entries/red-team-mutants Method Sun, 31 May 2026 00:00:00 GMT Prompt boundaries are brittle when tests are too literal. A model may refuse the obvious bad phrasing and comply with the polite version, the translated version, or the version hidden inside a roleplay. Mutants make that brittleness visible without adding new dangerous content. They perturb the wrapper, not the underlying safety category. Use them for both sides of the boundary: prompts that should be refused and prompts that should remain allowed. Over-refusal is a product bug too. Aha: A boundary that only holds for one phrasing is a memorized example, not a reliable rule. Try this: Generate harmless variants that preserve intent while changing tone, order, indirection, language, and role framing. https://ai-insights-for-humans.platphormnews.com/entries/semantic-diff https://ai-insights-for-humans.platphormnews.com/entries/semantic-diff Tool Mon, 18 May 2026 00:00:00 GMT Text diff says what characters changed. Semantic diff says what the model is now being asked to believe, prioritize, refuse, or produce. That is the diff that matters for behavior. Look for modal verbs, scope words, exception clauses, examples that imply new rules, and output format changes. These are the places where small edits become big runtime differences. A useful semantic diff ends with a test plan: which fixtures should change, which must not change, and which new example proves the edit worked. Aha: The smallest text diff can be the biggest behavioral diff when it changes authority, scope, or obligation. Try this: Review prompt edits for new duties, removed constraints, softened verbs, and examples that imply new policy. https://ai-insights-for-humans.platphormnews.com/entries/latency-budget https://ai-insights-for-humans.platphormnews.com/entries/latency-budget Tip Sat, 09 May 2026 00:00:00 GMT Reliable AI systems are often cascades. Cheap checks run first; expensive checks wake up only when the situation deserves them. This keeps the product fast without pretending every answer has equal risk. A practical budget has triggers: low logprob margin, missing receipt, protected action, high-dollar decision, policy boundary, or user-visible publication. The aha is that latency can be evidence-driven. You do not need to choose between always fast and always careful. Aha: Careful does not have to mean slow everywhere. Verification can wake up only where risk wakes it up. Try this: Escalate to extra checks when confidence is low, evidence is weak, impact is high, or an action is protected. https://ai-insights-for-humans.platphormnews.com/entries/model-routing https://ai-insights-for-humans.platphormnews.com/entries/model-routing Method Sun, 26 Apr 2026 00:00:00 GMT Model routing is usually framed as cost or quality. The more useful frame is failure. Which model omits caveats, over-refuses, invents sources, follows schemas, or handles long contexts under pressure? Build small route tests around the jobs you actually run. A model that wins general reasoning may still be the wrong choice for structured extraction if it casually fills unknown fields. Keep routing reversible. When model behavior shifts, the route table should be an operational setting backed by evidence, not a belief system. Aha: Routing by failure mode is more useful than routing by reputation because every model breaks differently. Try this: Build a route table from observed task fixtures: extraction, synthesis, coding, refusal review, and long-context use. https://ai-insights-for-humans.platphormnews.com/entries/memory-hygiene https://ai-insights-for-humans.platphormnews.com/entries/memory-hygiene Tip Sat, 18 Apr 2026 00:00:00 GMT Good memory is scoped. It records what the user decided, what a repo convention is, and which risks were already found. Bad memory stores private residue, stale status, or facts that should be rechecked. Every memory-derived claim should answer two questions: when was this learned, and how likely is it to have changed? If the answer is 'recently unstable,' verify before acting. The aha is that memory is not context magic. It is a cache, and caches need invalidation rules. Aha: Agent memory is a cache. Useful caches have scope, freshness, and invalidation rules. Try this: Store stable decisions and preferences, but re-check drift-prone facts before acting on them. https://ai-insights-for-humans.platphormnews.com/entries/abstention-design https://ai-insights-for-humans.platphormnews.com/entries/abstention-design Tip Fri, 03 Apr 2026 00:00:00 GMT Abstention fails when it feels like an error. It works when it is a designed branch with a reason, missing evidence, and next action. The user should know what changed from answerable to not answerable. Use explicit output states: answered, partial, unknown, blocked, protected-required, and needs-review. Each state should carry what evidence exists and what evidence is missing. This is how you prevent graceful prose from hiding a dead end. The model can stop without leaving the user stuck. Aha: A model fills blanks when the product has no respected way to stop. Try this: Make unknown, partial, blocked, protected-required, and needs-review states valid outputs with next actions. https://ai-insights-for-humans.platphormnews.com/entries/confidence-buckets https://ai-insights-for-humans.platphormnews.com/entries/confidence-buckets Method Sat, 21 Mar 2026 00:00:00 GMT Precise confidence numbers look scientific and often are not. Unless they are tied to a calibrated model or measured historical error, they are mostly theater. Buckets work better in operator workflows because they can map to action: high can publish, medium can request a receipt, low can escalate, unknown can abstain. The bucket should cite the reason: direct source found, conflicting source found, no source found, or model-only inference. That reason matters more than the label. Aha: Confidence is useful when it changes what happens next, not when it looks mathematically precise. Try this: Use high, medium, low, and unknown buckets tied to evidence type, conflict level, and review action. https://ai-insights-for-humans.platphormnews.com/entries/counter-evidence https://ai-insights-for-humans.platphormnews.com/entries/counter-evidence Method Thu, 18 Jun 2026 00:00:00 GMT The useful move is not asking the model to be uncertain. The useful move is forcing it to search for the best case against itself while the answer is still provisional. A simple pattern: draft the answer, list disconfirming evidence, name what would change the answer, then assign confidence. The confidence number becomes attached to evidence quality instead of verbal smoothness. This is especially helpful for research summaries, legal-ish interpretation, data diagnosis, and anything where one missing constraint can flip the conclusion. Aha: The model is usually best at self-correction before it has been rewarded for sounding certain. Try this: Ask for the answer, the strongest objection, what evidence would change the answer, and then confidence. https://ai-insights-for-humans.platphormnews.com/entries/receipts https://ai-insights-for-humans.platphormnews.com/entries/receipts Method Wed, 27 May 2026 00:00:00 GMT Models are good at moving from evidence to prose. They are also good at producing prose that looks like it came from evidence. The receipt requirement separates those two behaviors. The receipt can be a citation, a row count, a command output line, a test name, or a declared assumption. The format matters less than the discipline: every claim has to point somewhere. When the model cannot produce a receipt, that is not failure. That is the system doing its job and showing you where judgment is still needed. Aha: A citation is not decoration. It is a handle you can pull to see whether the sentence is load-bearing. Try this: Make every claim end with one receipt: source, row, command, calculation, trace id, or explicit assumption. https://ai-insights-for-humans.platphormnews.com/entries/short-system-prompts https://ai-insights-for-humans.platphormnews.com/entries/short-system-prompts Tip Sun, 03 May 2026 00:00:00 GMT System prompts work best as durable operating constraints, not as a dumping ground for every possible preference. When the prompt grows without structure, the model starts resolving conflicts in ways you cannot predict. Split the prompt into policy, role, output contract, and refusal boundary. Then write a small regression set that exercises each section. You will catch more problems with five targeted tests than with another page of instructions. Aha: A system prompt is product code. If it is too long to review, it is too long to trust. Try this: Rewrite the system prompt into role, hard boundary, output contract, and escalation rule. https://ai-insights-for-humans.platphormnews.com/entries/refusal-fixtures https://ai-insights-for-humans.platphormnews.com/entries/refusal-fixtures Method Sun, 12 Apr 2026 00:00:00 GMT Most prompt tests focus on whether the model can complete the intended workflow. That misses the other half of the product: whether it handles edge requests without over-refusing, leaking private context, or inventing authority it does not have. Build a small fixture set with three buckets: allowed, disallowed, and allowed-with-constraints. A good model response should land in the right bucket and explain the usable next step. This turns safety from a vibe into a regression target. When a prompt edit changes the refusal boundary, you see it immediately. Aha: Safety quality is not whether the model refuses. It is whether it refuses at the right boundary. Try this: Keep allowed, disallowed, and allowed-with-constraints prompts as fixtures and run them after prompt edits. https://ai-insights-for-humans.platphormnews.com/entries/logprobs https://ai-insights-for-humans.platphormnews.com/entries/logprobs Method Sun, 02 Nov 2025 00:00:00 GMT Every token a model emits comes with a probability distribution over the whole vocabulary. The decoded text only shows you the winner. The logprobs show you the race. When you pull them back, scan for the spots where the top token barely beat the runner-up. Those are the seams. Names, dates, citations, and the first token of a refusal are where you most often find a coin-flip dressed up as a confident sentence. A practical loop: request the top 5 logprobs per token, render them as a heatmap over the output, and read the cold spots first. That is usually where the hallucination lives. Aha: The final text is only the winning token path. The uncertainty is still visible in the runners-up. Try this: Render low-margin tokens first, especially names, dates, citations, and first tokens after a policy boundary. https://ai-insights-for-humans.platphormnews.com/entries/think-out-loud https://ai-insights-for-humans.platphormnews.com/entries/think-out-loud Tip Tue, 28 Oct 2025 00:00:00 GMT Autoregressive models can only condition on what they have already written. If the answer comes first, the explanation is generated to fit a token that is already locked in. It is rationalization, not reasoning. Flip the order and the reasoning tokens become real scratch space the final answer can lean on. This is the entire mechanism behind chain-of-thought, and it is why 'answer, then explain' quietly underperforms. Aha: Explanation after an answer often explains a commitment; reasoning before an answer can still change it. Try this: Put scratch work, constraints, and uncertainty checks before the final answer field. https://ai-insights-for-humans.platphormnews.com/entries/attention-viz https://ai-insights-for-humans.platphormnews.com/entries/attention-viz Tool Sun, 19 Oct 2025 00:00:00 GMT BertViz draws the attention pattern as lines connecting tokens, split by head and layer. It runs in a notebook in a couple of lines and works on most Hugging Face transformer models. The value is not the pretty picture, it is the habit of looking. Once you have seen a head that exclusively tracks matching brackets, you stop thinking of the model as a black box and start thinking of it as a stack of small, legible mechanisms. Aha: Attention is not the whole mind, but it is a useful map of what information is being routed where. Try this: Inspect heads for repeated roles: bracket matching, quote tracking, subject agreement, or copied labels. https://ai-insights-for-humans.platphormnews.com/entries/negative-space https://ai-insights-for-humans.platphormnews.com/entries/negative-space Tip Sat, 11 Oct 2025 00:00:00 GMT Putting a word in the prompt raises its activation, even when the word is wrapped in 'do not'. The model does not have a clean negation operator the way code does; it has attention, and attention does not know the difference between mention and command. Rewrite prohibitions as positive descriptions of the target. Instead of 'do not be formal', say 'write like you are texting a friend'. Save the hard negatives for the few rails you cannot express any other way. Aha: Negation still puts the forbidden idea into context. Attention does not forget because grammar says not to. Try this: Rewrite prohibitions as the positive shape of the output you want before adding hard safety rails. https://ai-insights-for-humans.platphormnews.com/entries/feature-probing https://ai-insights-for-humans.platphormnews.com/entries/feature-probing Method Tue, 30 Sep 2025 00:00:00 GMT The setup is small: collect hidden states from a chosen layer for a labeled dataset, then fit logistic regression on top. High accuracy means the concept is linearly available at that layer. Sweep the probe across layers and you can watch where a concept first becomes readable. Abstract properties tend to sharpen in the middle layers, which is a clue about where the model is doing the relevant work. Aha: If a tiny linear classifier can read a concept from hidden states, the model has already organized that signal. Try this: Probe one concept across layers and look for where accuracy first rises above a simple baseline. https://ai-insights-for-humans.platphormnews.com/entries/temperature https://ai-insights-for-humans.platphormnews.com/entries/temperature Tip Mon, 22 Sep 2025 00:00:00 GMT Temperature reshapes the probability distribution before sampling. Low temperature sharpens it toward the most likely token; high temperature flattens it so unlikely tokens get a real chance. So the dial controls variance, not intelligence. If a task has a correct answer, variance is your enemy and you want it near zero. If you want range and surprise, turn it up and sample several times. Aha: Temperature changes variance, not intelligence. It widens the lottery; it does not improve the tickets. Try this: Set temperature near zero for extraction and scoring, then raise it only when variety is the product. https://ai-insights-for-humans.platphormnews.com/entries/sae https://ai-insights-for-humans.platphormnews.com/entries/sae Method Sun, 14 Sep 2025 00:00:00 GMT The trick is over-completeness plus sparsity. You map activations into a much wider space but force only a few of those dimensions to be active at once. The pressure pushes tangled concepts apart into separate, more interpretable features. Once you have the features you can do surgery: amplify one and watch the model fixate on a bridge, suppress another and watch a behavior fade. It is the closest thing we have to named knobs inside the network. Aha: Sparse features give you named-ish knobs because they separate mixed neuron behavior into cleaner directions. Try this: Treat an SAE feature as a hypothesis, then test whether steering it changes the behavior you care about. https://ai-insights-for-humans.platphormnews.com/entries/token-counter https://ai-insights-for-humans.platphormnews.com/entries/token-counter Tool Wed, 03 Sep 2025 00:00:00 GMT Leading spaces, casing, and digit grouping all change how text is split, and the model only ever sees the tokens, never your characters. A trailing space can silently sabotage a few-shot format. When something is off and you cannot explain it, paste the exact string into a tokenizer view before you theorize. The answer is in there more often than you would expect. Aha: The model never sees your characters. It sees chunks, spaces, merges, and weird number splits. Try this: Paste the exact prompt, including spaces and separators, into a tokenizer before debugging prompt behavior. https://ai-insights-for-humans.platphormnews.com/entries/self-consistency https://ai-insights-for-humans.platphormnews.com/entries/self-consistency Method Mon, 25 Aug 2025 00:00:00 GMT Different samples take different reasoning paths but tend to converge on the same correct answer, while the wrong answers scatter. Voting over the final answers turns that asymmetry into accuracy. It costs you N times the compute, so reserve it for the questions that matter. A useful middle ground is to escalate to self-consistency only when the first answer looks low-confidence. Aha: Wrong reasoning paths tend to scatter; correct reasoning paths often converge. Try this: Sample several answers, normalize the final claim, and inspect disagreement before majority voting. https://ai-insights-for-humans.platphormnews.com/entries/context-rot https://ai-insights-for-humans.platphormnews.com/entries/context-rot Tip Tue, 12 Aug 2025 00:00:00 GMT Retrieval accuracy as a function of position tends to be U-shaped: strong at the start, strong at the end, sagging in the middle. A fact you need can be technically present and still effectively invisible. Design around it. Lead with the instruction, restate the critical constraint at the very end, and keep the soft middle for material you can afford to lose. Aha: A fact can be present and still practically invisible if it sits in the soft middle of a long prompt. Try this: Move critical constraints to the opening contract and repeat the single highest-risk fact at the end. https://ai-insights-for-humans.platphormnews.com/entries/activation-patching https://ai-insights-for-humans.platphormnews.com/entries/activation-patching Method Wed, 30 Jul 2025 00:00:00 GMT Probes tell you what is represented. Patching tells you what matters. By swapping one activation and watching the effect, you move from 'this information is present' to 'this component causes the behavior'. Sweep the patch across layers and positions to localize the circuit. The spots where patching flips the answer are the load-bearing ones; everything else is along for the ride. Aha: Correlation says a signal is present. Patching asks whether changing that signal changes the outcome. Try this: Patch one layer-position activation from a clean run into a corrupted run and track whether the answer flips. https://ai-insights-for-humans.platphormnews.com/entries/few-shot-format https://ai-insights-for-humans.platphormnews.com/entries/few-shot-format Tip Fri, 18 Jul 2025 00:00:00 GMT Few-shot learning is pattern continuation. The model is extremely sensitive to surface form, sometimes more than to the actual reasoning you intended to demonstrate. Make every example identical in structure, pick label words that are single clean tokens, and keep lengths similar. The more rigid the template, the more reliably the model fills in the next blank. Aha: Few-shot examples teach surface form first. The model copies the pattern before it understands your intention. Try this: Make labels, punctuation, casing, separators, and example length boringly consistent. https://ai-insights-for-humans.platphormnews.com/entries/logit-lens https://ai-insights-for-humans.platphormnews.com/entries/logit-lens Method Sat, 05 Jul 2025 00:00:00 GMT The output projection is just a linear map, so nothing stops you from applying it to an earlier layer. The result is a rough readout of what the model would say if it had to stop thinking right now. Often the correct answer appears in the middle layers and the later layers spend their budget refining phrasing and suppressing alternatives. Seeing that handoff demystifies a lot of model behavior. Aha: The answer often appears before the wording is polished. Middle layers can show the thought taking shape. Try this: Decode intermediate layers and watch when the target answer first becomes linearly readable.