diff --git a/docs/debugging.rst b/docs/debugging.rst
index 3dee6e5f545ae..e7c379a10700d 100644
--- a/docs/debugging.rst
+++ b/docs/debugging.rst
@@ -144,3 +144,87 @@ For example:
     def is_odd(x: ti.template()):
         ti.static_assert(x.data_type() == ti.i32, "is_odd() is only supported for i32")
         return x % 2 == 1
+
+
+Tips for debugging
+------------------
+
+Debugging a Taichi program can be hard even with the builtin tools above.
+Taichi developers are currently devoting themselves in improving error messages and warnings
+to help user find potential BUGs in their programs.
+
+Here we collected some common BUGs that one might encounter with a Taichi program:
+
+Static typing system
+++++++++++++++++++++
+
+Taichi pertend that it's a dynamical-typed language like Python, but it's actually a
+statically-typed language which will be translated into high performance CPU/GPU instructions.
+
+So the code behavior in Taichi-scope is actually very different from Python-scope!
+
+Type of a variable is simply **determined at its first initialization and never changes later**.
+
+Although static-type provides better performance and simplicity, but may leads to BUGs if
+users not distinguished Taichi-scope from Python-scope, e.g.:
+
+.. code-block:: python
+
+    @ti.kernel
+    def buggy():
+        ret = 0  # 0 is a integer, so `ret` is typed as int32
+        for i in range(3):
+            ret += 0.1 * i  # i32 += f32, the result is still stored in int32!
+        print(ret)  # will shows 0
+
+    buggy()
+
+The codes above shows a common BUG due to the limitation of the static-type system.
+The Taichi compiler should shows a warning like:
+
+.. code-block:: none
+
+    [W 06/27/20 21:43:51.853] [type_check.cpp:visit@66] [$19] Atomic add (float32 to int32) may lose precision.
+
+This means that it can not store a float32 result to int32.
+The solution is to type ``ret`` as float32 at the first place:
+
+.. code-block:: python
+
+    @ti.kernel
+    def not_buggy():
+        ret = 0.0  # 0 is a floating point number, so `ret` is typed as float32
+        for i in range(3):
+            ret += 0.1 * i  # f32 += f32, OK!!
+        print(ret)  # will shows 0.6
+
+    not_buggy()
+
+
+`@archibate <https://github.com/archibate>`_'s personal suggestion to prevent issues like this:
+
+* Recall the ``float ret = 0;`` in C/C++, always use ``ret = float(0)`` on **initialization**,
+  and ``ret = int(0)`` for integers. So that you are always clear of what type every variable.
+
+Advanced Optimization
++++++++++++++++++++++
+
+Taichi has a advanced optimization engine to make your Taichi kernel to be as fast as it could.
+But like the ``gcc -O3`` does, sometimes advanced optimization can leads to BUGs as it tried
+too hard, including runtime errors like:
+
+```RuntimeError: [verify.cpp:basic_verify@40] stmt 8 cannot have operand 7.```
+
+You may use ``ti.core.toggle_advance_optimization(False)`` to turn off advanced
+optimization and see if the issue still exists:
+
+.. code-block:: python
+
+    import taichi as ti
+
+    ti.init()
+    ti.core.toggle_advance_optimization()
+
+    ...
+
+If that fixed the issue, please report this BUG on `GitHub <https://github.com/taichi-dev/taichi/issues/new?labels=potential+bug&template=bug_report.md>`_ to help us improve, if you would like to.
diff --git a/python/taichi/__init__.py b/python/taichi/__init__.py
index 6c9a82c538c11..8cd8c17b36f63 100644
--- a/python/taichi/__init__.py
+++ b/python/taichi/__init__.py
@@ -7,7 +7,7 @@
 from taichi.misc import *
 from taichi.misc.gui import GUI
 from taichi.misc.np2ply import PLYWriter
-from taichi.misc.image import imread, imwrite, imshow
+from taichi.misc.image import *
 from taichi.misc.task import Task
 from taichi.misc.test import *
 from taichi.misc import settings as settings
diff --git a/python/taichi/lang/impl.py b/python/taichi/lang/impl.py
index 1f2d60488eea8..32ebfc8998564 100644
--- a/python/taichi/lang/impl.py
+++ b/python/taichi/lang/impl.py
@@ -195,6 +195,9 @@ def get_tape(self, loss=None):
     def sync(self):
         self.materialize()
         self.prog.synchronize()
+        # print's in kernel won't take effect until ti.sync(), discussion:
+        # https://github.com/taichi-dev/taichi/pull/1303#discussion_r444897102
+        print(taichi_lang_core.pop_python_print_buffer(), end='')
 
 
 pytaichi = PyTaichi()
diff --git a/taichi/backends/opengl/opengl_api.cpp b/taichi/backends/opengl/opengl_api.cpp
index 1bd9c94590cba..d466d1bc25ae4 100644
--- a/taichi/backends/opengl/opengl_api.cpp
+++ b/taichi/backends/opengl/opengl_api.cpp
@@ -5,6 +5,7 @@
 #include "taichi/program/kernel.h"
 #include "taichi/program/program.h"
 #include "taichi/util/environ_config.h"
+#include "taichi/python/print_buffer.h"
 #include "taichi/backends/opengl/runtime.h"
 
 #ifdef TI_WITH_OPENGL
@@ -487,7 +488,7 @@ struct CompiledProgram::Impl {
             TI_WARN("[glsl] Unexpected serialization type: {}, ignoring", type);
             break;
         };
-        std::cout << str;
+        py_cout << str;
       }
     }
     rt_buf->msg_count = 0;
diff --git a/taichi/python/export_misc.cpp b/taichi/python/export_misc.cpp
index 2c3c250692b6b..e48a3b8fb9d7c 100644
--- a/taichi/python/export_misc.cpp
+++ b/taichi/python/export_misc.cpp
@@ -7,6 +7,7 @@
 #include "taichi/common/task.h"
 #include "taichi/math/math.h"
 #include "taichi/python/exception.h"
+#include "taichi/python/print_buffer.h"
 #include "taichi/python/export.h"
 #include "taichi/system/benchmark.h"
 #include "taichi/system/profiler.h"
@@ -20,6 +21,8 @@
 
 TI_NAMESPACE_BEGIN
 
+PythonPrintBuffer py_cout;
+
 Config config_from_py_dict(py::dict &c) {
   Config config;
   for (auto item : c) {
@@ -153,6 +156,9 @@ void export_misc(py::module &m) {
     }
     printf("test was successful.\n");
   });
+  m.def("pop_python_print_buffer", []() {
+      return py_cout.pop_content();
+  });
   m.def("with_cuda", is_cuda_api_available);
   m.def("with_metal", taichi::lang::metal::is_metal_api_available);
   m.def("with_opengl", taichi::lang::opengl::is_opengl_api_available);
diff --git a/taichi/python/print_buffer.h b/taichi/python/print_buffer.h
new file mode 100644
index 0000000000000..857990cc04e50
--- /dev/null
+++ b/taichi/python/print_buffer.h
@@ -0,0 +1,25 @@
+#pragma once
+
+#include <sstream>
+
+TI_NAMESPACE_BEGIN
+
+struct PythonPrintBuffer {
+  /* holds kernel print result before switching back to python */
+  std::stringstream ss;
+
+  template <typename T>
+  PythonPrintBuffer &operator<<(const T &t) {
+    ss << t;
+    return *this;
+  }
+  std::string pop_content() {
+    auto ret = ss.str();
+    ss = std::stringstream();
+    return ret;
+  }
+};
+
+extern PythonPrintBuffer py_cout;
+
+TI_NAMESPACE_END