findings
104 insights derived from 4,656 threads, 208,799 messages, 20 users across 9 months.
synthesis
7 insights
EXECUTIVE SUMMARY
Top 5 findings and recommendations from analyzing 4,656 threads and 208,799 messages.
ULTIMATE SYNTHESIS
the ONE document. 4,656 threads. 208,799 messages. 20 users. 9 months. 48 insight files distilled.
MEGA SYNTHESIS
MEGA-SYNTHESIS: amp thread analysis
INDEX
insights index
DASHBOARD
> 4,656 threads analyzed | metrics derived from MEGA-SYNTHESIS
SYNTHESIS
compiled from 10 analysis documents spanning 4,281 threads (208,799 messages) across 20 users.
VERBOSE EXPLORER SUMMARY
verbose_explorer's amp summary
user behavior
7 insights
verbose_explorer improvement plan
verbose_explorer improvement plan
verbose_explorer specific
verbose_explorer's amp usage patterns: deep dive
power user behaviors
analysis of the three users with highest resolution rates: precision_pilot (82%), steady_navigator (67%), concise_commander (60.5%).
user comparison
Comparative analysis of verbose_explorer vs concise_commander showing why concise_commander achieves 60% vs 34% resolution.
user journey map
User progression from first thread to mastery over 5+ months of amp usage.
user onboarding
new to amp? this guide distills 4,656 threads into what actually matters.
user profiles
comprehensive analysis of 6 users with >50 threads from amp corpus.
patterns
89 insights
AGENTS MD FINAL
synthesized from analysis of 4,656 threads, 208,799 messages, 1,434 steering events, 2,050 approval events.
agent compliance
analysis of how often agent follows explicit user instructions across 500 threads (4656 available).
agent personality
synthesis from 4,656 threads, 23,262 messages, analyzing what agent behaviors succeed and fail.
agents md recommendations
synthesized from analysis of 4,281 threads, 208,799 messages, 901 steering events, 2,050 approval events.
anti patterns catalog
consolidated reference of agent anti-patterns from 4,656 thread analysis.
approval maximization
distilled from 4,656 threads, 208,799 messages. focus: what AGENT BEHAVIORS (not user behaviors) correlate with approval.
approval triggers
analysis of what assistant actions precede user APPROVAL messages.
assistant brevity
Medium-length responses (500-1k chars) get the best approval and resolution rates.
behavioral nudges
gentle interventions an agent can make during conversation to improve thread outcomes. derived from analysis of 4,656 amp threads.
best practices poster
Top 10 best practices for amp agent success based on analysis of 4,656 threads.
closing rituals
analysis of final user messages in 2,375 successfully closed threads (2,070 RESOLVED + 305 COMMITTED).
code quality signals
analysis of 4,656 threads for lint errors, type errors, test failures and their correlation with outcomes.
commit patterns
what distinguishes threads that reach `COMMITTED` status?
collaboration intensity
messages per hour calculated as `num_turns / (updated - created)` duration in hours.
common mistakes
derived from analysis of 4,656 amp threads. focuses on user-side patterns that correlate with lower resolution rates, higher steering, or frustrated o
comparative benchmarks
performance thresholds derived from 4,656 thread analysis. use to evaluate thread quality and user behavior.
complexity estimation
analysis of 4,281 threads to predict thread complexity (length, steering) from first message features.
context anchors
threads that explicitly reference prior work via:
context density
analysis of what constitutes dense, effective context in thread openers.
context window
estimated tokens derived from `char_len / 4` — rough approximation.
conversation dynamics
transition matrix built from 23,262 labeled messages across ~4,656 threads.
conversation templates
templates for common task types, derived from analysis of 4,656 threads. optimized for the patterns that correlate with resolution.
counter intuitive
patterns from 4,656 threads that contradict common assumptions about human-AI collaboration.
debug patterns
analysis of 678 threads containing debug, fix, or bug keywords.
domain expertise
analysis of unique vocabulary per user reveals distinct domain territories.
error analysis
analysis of error patterns in assistant messages from threads.db
early warning
analysis of 4,656 threads (14 FRUSTRATED, 1 STUCK) to identify earliest predictors of thread breakdown.
expletive analysis
analysis of user messages containing expletives (fuck, damn, wtf, hell, shit) across amp threads. investigates frustration triggers and patterns.
failure autopsy
analysis of 14 threads labeled FRUSTRATED. pattern extraction for breakdown points.
first message patterns
analysis of 4,281 threads with first user messages.
frustration signals
a production-ready detection system for identifying user frustration in amp threads, derived from analysis of 4,656 threads (14 FRUSTRATED, 1 STUCK).
git patterns
git patterns
golden examples
golden examples: 10 perfect threads
handoff network
handoff network analysis
handoff chains
handoff chains analysis
imperative analysis
imperative analysis: user message verbs and outcomes
implementation roadmap
implementation roadmap
instruction echo
instruction echo analysis
language patterns
language patterns: phrases that predict success vs failure
learning curves
learning curves: user evolution analysis
length analysis
thread length analysis by outcome
measurement framework
MEASUREMENT FRAMEWORK
memorable quotes
memorable quotes from frustrated threads
message brevity
analysis of 208,799 messages across 4,281 threads.
midnight analysis
deep dive on late night threads which showed 60.4% resolution rate—nearly double the evening rate.
multi file edits
analysis of 3,312 threads with file editing operations (71% of 4,656 total threads).
negative examples
analysis of threads with FRUSTRATED status or high steering counts (>5). documents what went wrong and lessons le@swift_solverd.
open questions
the analysis is extensive (4,656 threads, 208,799 messages, ~100 insight files) but significant gaps remain. organized by severity.
opening words
analysis of first 3 words from 4,281 user thread openers. correlates opening patterns with thread outcomes (message count, tool usage).
oracle timing
analyzed oracle mentions in assistant messages across 757 threads that invoked the oracle tool.
persistence analysis
what distinguishes threads that persist through difficulty vs those that abandon?
plan vs execute
analyzed 3488 threads for whether they start with planning/discussion or jump straight to execution.
positive examples
analysis of 20 best-performing threads: high-outcome (COMMITTED), zero steering interventions.
pre thread checklist
simple yes/no checklist before starting an amp thread.
prompting styles
analyzed 4281 threads with first user messages.
question analysis
analysis of 4,600 QUESTION-labeled messages across threads.
quick wins
ranked by effect size × ease of implementation.
recovery patterns
analysis of 552 threads that received STEERING corrections but ended RESOLVED.
refactoring patterns
analysis of 245 threads containing refactor, migrate, or upgrade in titles.
retro questions
structured questions for teams to discuss in retrospectives, organized by theme. each question is grounded in analysis of 4,656 threads.
sentence starters
extracted first 5 words of user openers, grouped by thread outcome.
shortcut patterns
analysis of high-steering threads to identify agent behaviors users actively reject.
signal strength ranking
predictive power for thread resolution, ranked by effect size and reliability.
skill recommendations
based on analysis of 4,656 threads across amp users.
skill usage
searched for load the skill and use the skill patterns across 4656 threads in threads.db.
spawn vs inline
analysis of 4,656 amp threads comparing threads that spawn subtasks via Task tool or tmux versus threads that stay inline.
steering deep dive
analysis of 1,434 steering messages across 23,262 user messages in the corpus.
steering taxonomy
complete classification of steering behaviors observed in 1,434 steering messages across 23,262 user messages.
success patterns
analysis of 2375 successful threads (RESOLVED + COMMITTED) vs 14 frustrated threads.
task delegation
analysis of 4,656 threads from amp usage data.
team patterns
extracted from 4,656 threads across 18 users.
testing patterns
analysis of test-related thread patterns from threads.db.
thread flow
analysis of 4,281 threads (208,799 messages) examining structural patterns that correlate with outcomes.
thread grading rubric
grading system for amp thread quality. A-F scale derived from 4,656 thread analysis.
thread lifecycle
analysis of 4,656 threads mapping the typical lifecycle of successful vs failed threads.
thread titles
analysis of 4,656 thread titles across outcome categories.
threading depth
Analysis of thread spawning patterns, chain depths up to 72, and orphan rates.
time analysis
analysis of 4,656 threads spanning 2025-05-12 to 2026-01-08 (~8 months)
tomorrow actions
prioritized for immediate impact. do these before anything else.
tool chains
extracted from 4,656 threads, 168,640 tool sequences.
tool patterns
analysis of 185,537 assistant messages across 4,259 threads.
topic clusters
keyword clustering on 4656 threads (excluding Untitled).
training curriculum
Evidence-based 4-week onboarding program for amp users distilled from thread analysis.
verification gates
threads that verify before declaring done (test runs, reviews, build checks) vs threads that don't.
vocabulary analysis
Extracted from user messages in threads.db.
web research human ai
web research synthesis on effective prompting styles, correction patterns, and how users learn to work with AI.
web research nlp
research compiled from academic sources and industry practices.
web research personality
web research findings for amp thread analysis project.
weekend analysis
investigating why weekend threads show +5.2pp higher resolution rates (48.9% vs 43.7%)
meta
1 insights