diff --git a/.gitignore b/.gitignore
index 773bb7b..b6931c6 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,5 @@
 *.dot
 *.png
-*dist_seq
+*.pdf
+*_seq
 *splitting_tree
diff --git a/CMakeLists.txt b/CMakeLists.txt
index f6e95ba..7e55e10 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 2.8)
 
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++1y")
 
-find_package(Boost REQUIRED COMPONENTS program_options filesystem system serialization)
+find_package(Boost REQUIRED COMPONENTS iostreams program_options filesystem system serialization)
 include_directories(SYSTEM ${Boost_INCLUDE_DIRS})
 set(libs ${libs} ${Boost_LIBRARIES})
 
diff --git a/lib/create_splitting_tree.cpp b/lib/create_splitting_tree.cpp
index abc3311..211ab11 100644
--- a/lib/create_splitting_tree.cpp
+++ b/lib/create_splitting_tree.cpp
@@ -15,7 +15,7 @@ std::vector<T> concat(std::vector<T> const & l, std::vector<T> const & r){
 	return ret;
 }
 
-result create_splitting_tree(const Mealy& g){
+result create_splitting_tree(const Mealy& g, options opt){
 	const auto N = g.graph.size();
 	const auto P = g.input_indices.size();
 	const auto Q = g.output_indices.size();
@@ -47,7 +47,9 @@ result create_splitting_tree(const Mealy& g){
 
 		assert(boom.states.size() == accumulate(begin(boom.children), end(boom.children), 0, [](auto l, auto r) { return l + r.states.size(); }));
 	};
-	const auto is_valid = [N, &g](auto blocks, auto symbol){
+	const auto is_valid = [N, opt, &g](auto blocks, auto symbol){
+		if(!opt.check_validity) return true;
+
 		for(auto && block : blocks) {
 			const auto new_blocks = partition_(begin(block), end(block), [symbol, &g](state state){
 				return apply(g, state, symbol).to.base();
diff --git a/lib/create_splitting_tree.hpp b/lib/create_splitting_tree.hpp
index af76af6..213393a 100644
--- a/lib/create_splitting_tree.hpp
+++ b/lib/create_splitting_tree.hpp
@@ -23,4 +23,11 @@ struct result {
 	bool is_complete;
 };
 
-result create_splitting_tree(Mealy const & m);
+struct options {
+	bool check_validity = true;
+};
+
+constexpr options with_validity_check{true};
+constexpr options without_validity_check{false};
+
+result create_splitting_tree(Mealy const & m, options opt);
diff --git a/lib/logging.hpp b/lib/logging.hpp
index cbb6692..cbcd1d5 100644
--- a/lib/logging.hpp
+++ b/lib/logging.hpp
@@ -1,10 +1,36 @@
 #pragma once
 
-#include <mutex>
+#include <chrono>
+#include <iostream>
 
-// Works particularly nice with lambda's, as they give naturally unique types :)
-template <typename F>
-void fire_once(F && f){
-	static std::once_flag flag;
-	std::call_once(flag, f);
-}
+struct timer{
+	using clock = std::chrono::high_resolution_clock;
+	using time = std::chrono::time_point<clock>;
+	using seconds = std::chrono::duration<double>;
+
+	std::string name;
+	time begin;
+	bool active = true;
+
+	timer(std::string name)
+	: name(name)
+	, begin(clock::now())
+	{
+		std::cerr << name << std::endl;
+	}
+
+	void stop(){
+		if(!active) return;
+		time end = clock::now();
+		std::cerr << "* " << from_duration(end - begin) << '\t' << name << std::endl;
+		active = false;
+	}
+
+	~timer(){
+		stop();
+	}
+
+	static double from_duration(seconds s){
+		return s.count();
+	}
+};
diff --git a/scripts.md b/scripts.md
new file mode 100644
index 0000000..8123016
--- /dev/null
+++ b/scripts.md
@@ -0,0 +1,46 @@
+all
+===
+
+
+for f in *.dot; do ../build/main $f; done
+for f in *splitting_tree; do sed -i "" -e 's/label="........................................*"/label="truncated"/g' $f; done
+for f in *dist_seq; do sed -i "" -e 's/label="........................................*"/label="truncated"/g' $f; done
+dot -O -Tpng -Goverlap=false *dist_seq
+dot -O -Tpng -Goverlap=false *splitting_tree
+
+
+graphs
+======
+
+dot -O -Tpng -Goverlap=false *dist_seq
+dot -O -Tpng -Goverlap=false *splitting_tree
+
+neato -O -Tpng -Goverlap=false *.dot
+
+
+truncation
+==========
+
+sed 's/label="........................................*"/label="truncated"/g' esm-manual-controller.dot.splitting_tree.dot > esm-manual-controller.dot.splitting_tree_truncated.dot
+
+for f in *splitting_tree; do echo $f; sed -i "" -e 's/label="........................................*"/label="truncated"/g' $f; done
+for f in *dist_seq; do echo $f; sed -i "" -e 's/label="........................................*"/label="truncated"/g' $f; done
+
+
+cleaning
+========
+
+sed 's/\[label.*<td>I/ \[label="/g;s/<\/td>.*<td>O/ \/ /g;s/<\/td>.*]/"\];/g' 
+
+
+not needed ctor
+===============
+
+partition_refine(Blocks other_blocks){
+	auto beg = elements.begin();
+	for(auto && block : other_blocks){
+		std::copy(block.begin(), block.end(), std::back_inserter(elements));
+		blocks.insert(blocks.end(), {beg, std::prev(elements.end())});
+		beg = elements.end();
+	}
+}
diff --git a/src/conf.cpp b/src/conf.cpp
new file mode 100644
index 0000000..99e1c0a
--- /dev/null
+++ b/src/conf.cpp
@@ -0,0 +1,50 @@
+#include <mealy.hpp>
+#include <read_mealy_from_dot.hpp>
+
+#include <boost/iostreams/device/file_descriptor.hpp>
+#include <boost/iostreams/filtering_stream.hpp>
+#include <boost/iostreams/filter/gzip.hpp>
+
+#include <fstream>
+#include <iostream>
+#include <string>
+
+using namespace std;
+
+int main(int argc, char *argv[]){
+	if(argc != 3) return 37;
+
+	const string m_filename = argv[1];
+	const string c_filename = argv[2];
+
+	ifstream m_file(m_filename);
+	boost::iostreams::filtering_istream c_file;
+	c_file.push(boost::iostreams::gzip_decompressor());
+	c_file.push(boost::iostreams::file_descriptor_source(c_filename));
+
+	const auto machine = read_mealy_from_dot(m_file);
+
+	string in;
+	string out;
+
+	state s = 0;
+	size_t count = 0;
+	while(c_file >> in >> out){
+		const auto i = machine.input_indices.at(in);
+		const auto o = machine.output_indices.at(out);
+
+		const auto ret = apply(machine, s, i);
+		if(ret.output != o){
+			cout << "conformance fail" << endl;
+			cout << ret.output << " != " << o << endl;
+			cout << "at index " << count << endl;
+			return 1;
+		}
+
+		s = ret.to;
+		count++;
+	}
+
+	cout << "conformance succes " << count << endl;
+}
+
diff --git a/src/main.cpp b/src/main.cpp
index 3dc0e59..1983c5f 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -2,6 +2,11 @@
 #include <create_splitting_tree.hpp>
 #include <read_mealy_from_dot.hpp>
 #include <write_tree_to_dot.hpp>
+#include <logging.hpp>
+
+#include <boost/iostreams/device/file_descriptor.hpp>
+#include <boost/iostreams/filtering_stream.hpp>
+#include <boost/iostreams/filter/gzip.hpp>
 
 #include <cassert>
 #include <iostream>
@@ -10,59 +15,267 @@
 #include <functional>
 #include <vector>
 #include <utility>
+#include <fstream>
 
 using namespace std;
 
+template <typename T>
+vector<string> create_reverse_map(map<string, T> const & indices){
+	vector<string> ret(indices.size());
+	for(auto&& p : indices){
+		ret[p.second.base()] = p.first;
+	}
+	return ret;
+}
+
+auto bfs(Mealy const & machine, state s){
+	vector<bool> visited(machine.graph_size, false);
+	vector<vector<input>> words(machine.graph_size);
+
+	queue<state> work;
+	work.push(s);
+	while(!work.empty()){
+		const auto u = work.front();
+		work.pop();
+
+		if(visited[u.base()]) continue;
+
+		visited[u.base()] = true;
+
+		for(input i = 0; i < machine.input_size; ++i){
+			const auto v = apply(machine, u, i).to;
+			if(visited[v.base()]) continue;
+
+			words[v.base()] = words[u.base()];
+			words[v.base()].push_back(i);
+			work.push(v);
+		}
+	}
+
+	return words;
+}
+
 int main(int argc, char *argv[]){
 	if(argc != 2) return 1;
 	const string filename = argv[1];
 
-	cerr << "* Reading file " << filename << "\n";
-	const auto machine = read_mealy_from_dot(filename);
-	assert(is_complete(machine));
-	cerr << "\tdone\n";
+	const auto machine = [&]{
+		timer t("reading file " + filename);
+		return read_mealy_from_dot(filename);
+	}();
 
-	cerr << "* Starting Lee & Yannakakis I\n";
-	const auto splitting_tree = create_splitting_tree(machine);
-	cerr << "\tdone\n";
+	const auto splitting_tree_hopcroft = [&]{
+		timer t("creating hopcroft splitting tree");
+		return create_splitting_tree(machine, without_validity_check);
+	}();
 
-	cerr << "* Write splitting tree\n";
-	const string tree_filename = splitting_tree.is_complete ? (filename + ".splitting_tree") : (filename + ".incomplete_splitting_tree");
-	write_splitting_tree_to_dot(splitting_tree.root, tree_filename);
-	cerr << "\tdone\n";
+	const auto all_pair_seperating_sequences = [&]{
+		timer t("gathering all seperating sequences");
 
-	cerr << "* Lee and Yannakaki II\n";
-	const auto distinguishing_sequence = create_adaptive_distinguishing_sequence(splitting_tree);
-	cerr << "\tdone\n";
+		vector<vector<vector<input>>> all_pair_seperating_sequences(machine.graph_size, vector<vector<input>>(machine.graph_size));
 
-	cerr << "* Write dist sequence\n";
-	const string dseq_filename = splitting_tree.is_complete ? (filename + ".dist_seq") : (filename + ".incomplete_dist_seq");
-	write_adaptive_distinguishing_sequence_to_dot(distinguishing_sequence.sequence, dseq_filename);
-	cerr << "\tdone\n" << endl;
+		queue<reference_wrapper<const splijtboom>> work;
+		work.push(splitting_tree_hopcroft.root);
 
-	vector<vector<input>> uios(splitting_tree.root.states.size());
-	stack<pair<vector<input>, reference_wrapper<const dist_seq>>> work;
-	work.push({{}, distinguishing_sequence.sequence});
+		// total complexity is O(n^2), as we're visiting each pair only once :)
+		while(!work.empty()){
+			const splijtboom & node = work.front();
+			work.pop();
 
-	while(!work.empty()){
-		auto word = work.top().first;
-		const dist_seq & node = work.top().second;
-		work.pop();
+			auto it = begin(node.children);
+			auto ed = end(node.children);
 
-		if(node.CI.size() == 1){
-			const auto state = node.CI[0].second;
-			uios[state.base()] = word;
-			continue;
+			while(it != ed){
+				auto jt = next(it);
+				while(jt != ed){
+					for(auto && s : it->states){
+						for(auto && t : jt->states){
+							assert(all_pair_seperating_sequences[t.base()][s.base()].empty());
+							assert(all_pair_seperating_sequences[s.base()][t.base()].empty());
+							all_pair_seperating_sequences[t.base()][s.base()] = node.seperator;
+							all_pair_seperating_sequences[s.base()][t.base()] = node.seperator;
+						}
+					}
+					jt++;
+				}
+				it++;
+			}
+
+			for(auto && c : node.children){
+				work.push(c);
+			}
 		}
 
-		for(auto && i : node.word)
-			word.push_back(i);
+		for(size_t i = 0; i < machine.graph_size; ++i){
+			for(size_t j = 0; j < machine.graph_size; ++j){
+				if(i == j) continue;
+				assert(!all_pair_seperating_sequences[i][j].empty());
+			}
+		}
 
-		for(auto && c : node.children)
-			work.push({word, c});
+		return all_pair_seperating_sequences;
+	}();
+
+	const auto splitting_tree = [&]{
+		timer t("Lee & Yannakakis I");
+		return create_splitting_tree(machine, with_validity_check);
+	}();
+
+	if(false){
+		timer t("writing splitting tree");
+		const string tree_filename = splitting_tree.is_complete ? (filename + ".splitting_tree") : (filename + ".incomplete_splitting_tree");
+		write_splitting_tree_to_dot(splitting_tree.root, tree_filename);
 	}
 
-	size_t uio_count = count_if(begin(uios), end(uios), [](auto && u){ return !u.empty(); });
-	cout << uio_count << " / " << uios.size() << endl;
+	const auto distinguishing_sequence = [&]{
+		timer t("Lee & Yannakakis II");
+		return create_adaptive_distinguishing_sequence(splitting_tree);
+	}();
+
+	if(false){
+		timer t("writing dist sequence");
+		const string dseq_filename = splitting_tree.is_complete ? (filename + ".dist_seq") : (filename + ".incomplete_dist_seq");
+		write_adaptive_distinguishing_sequence_to_dot(distinguishing_sequence.sequence, dseq_filename);
+	}
+
+	const auto seperating_family = [&]{
+		timer t("making seperating family");
+		using Word = vector<input>;
+		using SepSet = vector<Word>;
+		vector<SepSet> seperating_family(machine.graph_size);
+
+		stack<pair<vector<input>, reference_wrapper<const dist_seq>>> work;
+		work.push({{}, distinguishing_sequence.sequence});
+
+		while(!work.empty()){
+			auto word = work.top().first;
+			const dist_seq & node = work.top().second;
+			work.pop();
+
+			if(node.children.empty()){
+				// add sequence to this leave
+				for(auto && p : node.CI){
+					const auto state = p.second;
+					seperating_family[state.base()].push_back(word);
+				}
+
+				// if the leaf is not a singleton, we need the all_pair seperating seqs
+				for(auto && p : node.CI){
+					for(auto && q : node.CI){
+						const auto s = p.second;
+						const auto t = q.second;
+						if(s == t) continue;
+						seperating_family[s.base()].push_back(all_pair_seperating_sequences[s.base()][t.base()]);
+					}
+				}
+
+				continue;
+			}
+
+			for(auto && i : node.word)
+				word.push_back(i);
+
+			for(auto && c : node.children)
+				work.push({word, c});
+		}
+
+		return seperating_family;
+	}();
+
+	const auto inputs = create_reverse_map(machine.input_indices);
+	const auto outputs = create_reverse_map(machine.output_indices);
+	const auto print_uio = [&](auto const & word, auto & out, state s) -> auto & {
+		for(auto && i : word){
+			const auto o = apply(machine, s, i);
+			s = o.to;
+
+			out << inputs[i.base()] << ' ' << outputs[o.output.base()] << '\n';
+		}
+		return out;
+	};
+
+	const auto transfer_sequences = [&]{
+		timer t("determining transfer sequences");
+		vector<vector<vector<input>>> transfer_sequences(machine.graph_size);
+		for(state s = 0; s < machine.graph_size; ++s){
+			transfer_sequences[s.base()] = bfs(machine, s);
+		}
+		return transfer_sequences;
+	}();
+
+	const auto short_checking_seq = [&]{
+		timer t("making short checking seq");
+		vector<input> big_seq;
+		state from = 0;
+		for(state s = from; s < machine.graph_size; ++s){
+			for(const auto & seq : seperating_family[s.base()]){
+				copy(begin(seq), end(seq), back_inserter(big_seq));
+				from = apply(machine, s, begin(seq), end(seq)).to;
+
+				const auto to = s;
+				if(from == to) continue;
+
+				const auto transfer = transfer_sequences[from.base()][to.base()];
+				copy(begin(transfer), end(transfer), back_inserter(big_seq));
+			}
+
+			const auto to = s+1;
+			if(from == to) continue;
+
+			const auto transfer = transfer_sequences[from.base()][to.base()];
+			copy(begin(transfer), end(transfer), back_inserter(big_seq));
+		}
+
+		return big_seq;
+	}();
+
+	{
+		timer t("writing short checking seq");
+		const string uios_filename = filename + ".short_check_seq";
+
+		boost::iostreams::filtering_ostream out;
+		out.push(boost::iostreams::gzip_compressor());
+		out.push(boost::iostreams::file_descriptor_sink(uios_filename));
+
+		print_uio(short_checking_seq, out, 0);
+	}
+
+	const auto long_checking_seq = [&]{
+		timer t("making long checking seq");
+		vector<input> big_seq;
+		state from = 0;
+		for(state s = from; s < machine.graph_size; ++s){
+			for(input i = 0; i < machine.input_size; ++i){
+				const auto t = apply(machine, s, i).to;
+
+				for(auto && seq : seperating_family[t.base()]){
+					if(from != s){
+						const auto transfer = transfer_sequences[from.base()][s.base()];
+						copy(begin(transfer), end(transfer), back_inserter(big_seq));
+						from = s;
+					}
+
+					big_seq.push_back(i);
+					from = t;
+
+					copy(begin(seq), end(seq), back_inserter(big_seq));
+					from = apply(machine, from, begin(seq), end(seq)).to;
+				}
+			}
+		}
+
+		return big_seq;
+	}();
+
+	{
+		timer t("writing long checking seq");
+		const string uios_filename = filename + ".full_check_seq";
+
+		boost::iostreams::filtering_ostream out;
+		out.push(boost::iostreams::gzip_compressor());
+		out.push(boost::iostreams::file_descriptor_sink(uios_filename));
+
+		print_uio(long_checking_seq, out, 0);
+	}
 }
 
diff --git a/src/stats.cpp b/src/stats.cpp
new file mode 100644
index 0000000..bb60817
--- /dev/null
+++ b/src/stats.cpp
@@ -0,0 +1,22 @@
+#include <read_mealy_from_dot.hpp>
+#include <mealy.hpp>
+
+#include <string>
+#include <fstream>
+#include <iostream>
+
+using namespace std;
+
+int main(int argc, char *argv[]){
+	if(argc != 2) return 37;
+
+	const auto filename = argv[1];
+	ifstream file(filename);
+	const auto machine = read_mealy_from_dot(file);
+
+	cout << "machine " << filename << " has\n";
+	cout << '\t' << machine.graph_size << " states\n";
+	cout << '\t' << machine.input_size << " inputs\n";
+	cout << '\t' << machine.output_size << " outputs\n";
+}
+