diff --git a/docs/src/concepts/1_screen.md b/docs/src/concepts/1_screen.md
index 2d9cd54b6..97472d1cd 100644
--- a/docs/src/concepts/1_screen.md
+++ b/docs/src/concepts/1_screen.md
@@ -127,10 +127,6 @@ collect(zip(
 This may be easily used for e.g. scrutinizing all possible reaction pairs, to
 find the ones that are redundant and not.
 
-!!! tip "Notebook available"
-    A notebook is available that demonstrates
-    [the screening on a larger scale](../notebooks/6_screening.md).
-
 There are many other variant "specifications" to choose from. You may use
 [`with_added_reactions`](@ref), [`with_removed_reactions`](@ref),
 [`with_removed_metabolites`](@ref), and others. Function reference contains a
diff --git a/docs/src/concepts/2_modifications.md b/docs/src/concepts/2_modifications.md
index e53bbac99..26195afcf 100644
--- a/docs/src/concepts/2_modifications.md
+++ b/docs/src/concepts/2_modifications.md
@@ -10,17 +10,16 @@ list callbacks that do the changes to the prepared optimization model.
 The callbacks available in COBREXA.jl include functions that may help with
 tuning the optimizer, or change the raw values in the linear model, such as:
 
-- [`change_objective_value`](@ref) and [`change_objective`](@ref)
-- [`change_constraint`](@ref)
+- [`change_constraint`](@ref) and [`change_objective`](@ref)
 - [`change_sense`](@ref), [`change_optimizer`](@ref), [`change_optimizer_attribute`](@ref)
 - [`silence`](@ref)
 - [`knockout`](@ref), [`add_crowding_constraints`](@ref)
 
-Compared to the [variant system](TODO) and the [model wrappers](TODO),
-optimizer modifications are slightly more powerful (they can do anything they
-want with the optimizer!), but do not compose well -- it is very easy to break
-the semantics of the model or erase the previous changes by carelessly adding
-the modifications.
+Compared to the [variant system](1_screen.md) and the [model
+wrappers](4_wrappers.md), optimizer modifications are slightly more powerful
+(they can do anything they want with the optimizer!), but do not compose well
+-- it is very easy to break the semantics of the model or erase the previous
+changes by carelessly adding the modifications.
 
 Here, we show how to construct the modifications. Their semantics is similar to
 the [variant-generating functions](1_screen.md), which receive a model (of type
diff --git a/docs/src/concepts/3_custom_models.md b/docs/src/concepts/3_custom_models.md
index 4a923d960..957b35552 100644
--- a/docs/src/concepts/3_custom_models.md
+++ b/docs/src/concepts/3_custom_models.md
@@ -29,10 +29,6 @@ making small changes to the model using the modifications system, with
 functions such as [`with_added_reactions`](@ref) and
 [`with_changed_bound`](@ref).
 
-!!! tip "Notebook available"
-    A better example of using a custom model structure is available
-    [in a separate notebook](../notebooks/8_custom_model.md).
-
 ## Writing the generic accessors
 
 Let's write a data structure that represents a very small model that contains N
diff --git a/docs/src/examples/15_gecko.jl b/docs/src/examples/15_gecko.jl
index a0fc89510..01e5a1a03 100644
--- a/docs/src/examples/15_gecko.jl
+++ b/docs/src/examples/15_gecko.jl
@@ -4,20 +4,20 @@
 # the cell to respect many known parameters, measured by proteomics and other
 # methods.
 #
-# The original description from GECKO is by:
-# Sánchez, B.J., Zhang, C., Nilsson, A., Lahtvee, P.J., Kerkhoven, E.J. and
-# Nielsen, J., 2017. "Improving the phenotype predictions of a yeast
-# genome‐scale metabolic model by incorporating enzymatic constraints."
-# *Molecular systems biology*, 13(8), p.935.
+# The original description from GECKO is by: Sánchez, B.J., Zhang, C., Nilsson,
+# A., Lahtvee, P.J., Kerkhoven, E.J. and Nielsen, J., 2017. "Improving the
+# phenotype predictions of a yeast genome‐scale metabolic model by
+# incorporating enzymatic constraints." *Molecular systems biology*, 13(8),
+# p.935.
 #
 # The analysis method and implementation in COBREXA is principially similar to
-# [sMOMENT](14_smoment), but GECKO is able to process and represent much larger scale
-# of the constraints -- mainly, it supports multiple isozymes for each
-# reaction, and the isozymes can be grouped into "enzyme mass groups" to
+# [sMOMENT](14_smoment.md), but GECKO is able to process and represent much
+# larger scale of the constraints -- mainly, it supports multiple isozymes for
+# each reaction, and the isozymes can be grouped into "enzyme mass groups" to
 # simplify interpretation of data from proteomics.
 
 # For demonstration, we will generate artificial random data in a way similar
-# to the [sMOMENT example](14_smoment):
+# to the [sMOMENT example](14_smoment.md):
 
 !isfile("e_coli_core.json") &&
     download("http://bigg.ucsd.edu/static/models/e_coli_core.json", "e_coli_core.json")