What's New

Everything added to VizLearn, newest first.

Updated 15 September 2026

This is the written record of what VizLearn has become, in the order it happened. Every entry below is either a set of modules that went live on a particular day, or a change to how the site itself works — a new kind of visualisation, a fix to something that was misleading, a rewrite of a page that was not carrying its weight.

Where the dates come from

The module entries are not maintained by hand. Each module's first commit in the repository is what dates it, and this page groups modules by that date. That means a module appears here the moment it actually ships, and the list cannot quietly drift out of step with the site. The same date drives the published date on the module's own page, so the two can never disagree.

Everything that is not a module — the features, the corrections, the structural changes — is written by hand, because nothing in the repository knows why a change was made. Those entries say what changed and, where it matters, what was wrong before.

What counts as news

Modules committed before 1 August 2026 are not listed. They are not new; they are what the site was when it started, and putting a hundred and eighty of them under a single date would bury everything that came after. If you are looking for those, the track pages list every module in the order they are meant to be read, and the concept map shows how the tracks depend on one another.

Reading a release

Each block names its date, says which tracks grew and by how much, and then lists the individual modules with the track each one belongs to. Every title is a link straight to the module. Where a release also changed how the site behaves, those notes sit above the module list.

471 modules added since 1 August 2026

15 September 2026

New modules

Ten new modules, all of them in the Interview Questions track.

Binary search on the answer, not the arrayInterview Questions Climbing stairs: the DP that is FibonacciInterview Questions Coin change, and why greedy is wrongInterview Questions Find the median of a data streamInterview Questions First and last position of a targetInterview Questions House robber: the 0-1 choiceInterview Questions Longest increasing subsequence, twiceInterview Questions Merge k sorted sequencesInterview Questions Top k frequent elementsInterview Questions Write binary search without an off-by-oneInterview Questions

14 September 2026

New modules

Four new modules, all of them in the Interview Questions track.

Daily temperatures, and the monotonic stackInterview Questions Design a stack that reports its minimum in O(1)Interview Questions Evaluate reverse Polish notationInterview Questions Implement a queue using two stacksInterview Questions

12 September 2026

New modules

Eight new modules, all of them in the Interview Questions track.

Shallow copy vs deep copyInterview Questions What does `with` guarantee when the body raises?Interview Questions What does `yield` actually do?Interview Questions What does a decorator actually replace?Interview Questions Why do all these functions return the same value?Interview Questions Why does this default argument remember?Interview Questions Why is a comprehension faster than the same loop?Interview Questions `is` vs `==`, and why 257 is not 257Interview Questions

11 September 2026

New modules

35 new modules: eight in Generative AI & LLMs, seven in Computer Vision, seven in Concurrency, five in Async Python, four in Deep Learning and four in Natural Language Processing.

1 September 2026

New modules

14 new modules, all of them in the scikit-learn track.

A Complete Workflowscikit-learn Class Imbalancescikit-learn Decision Treesscikit-learn Different Columns, Different Treatmentscikit-learn Encoding Categoriesscikit-learn Hyperparameter Searchscikit-learn Leakage in a scikit-learn Workflowscikit-learn Missing Valuesscikit-learn Overfitting and Underfittingscikit-learn Pipelines in scikit-learnscikit-learn Probabilities and Thresholdsscikit-learn Random Forestsscikit-learn Scaling Featuresscikit-learn k-Means Clusteringscikit-learn

31 August 2026

New modules

57 new modules: 26 in pandas, 22 in matplotlib and nine in scikit-learn.

Adding and Removing Columnspandas Annotating a Plotmatplotlib Bar Chartsmatplotlib Box and Violin Plotsmatplotlib Choosing a Chartmatplotlib Classification Metricsscikit-learn Colour and Colormapsmatplotlib Common Mistakesmatplotlib Counting and Binningpandas Creating DataFramespandas Cross-Validationscikit-learn Dates and Timespandas Dates on an Axismatplotlib Dtypes and Memorypandas Duplicatespandas Error Bars and Bandsmatplotlib Figure and Axesmatplotlib Filtering Rowspandas Histograms and Distributionsmatplotlib Images and Heatmapsmatplotlib Labels, Titles and Legendsmatplotlib Layout and Spacingmatplotlib Limits, Ticks and Scalesmatplotlib Line Plotsmatplotlib Linear Regressionscikit-learn Loading and Shaping Datascikit-learn Logistic Regressionscikit-learn Looking at Datapandas Method Chainingpandas Missing Datapandas Performancepandas Performance and Large Datamatplotlib Plotting from pandasmatplotlib Reading and Writing Filespandas Regression Metricsscikit-learn Reshaping: pivot and meltpandas Saving Figuresmatplotlib Scatter Plotsmatplotlib Sorting and Rankingpandas Splitting Train and Testscikit-learn Stacking Frames with concatpandas Styles and rcParamsmatplotlib Subplotsmatplotlib Text Columnspandas The Confusion Matrixscikit-learn The Copy Warningpandas The Indexpandas The MultiIndexpandas Time Seriespandas Twin and Secondary Axesmatplotlib What pandas Is Forpandas apply, map and Vectorisingpandas fit, predict, transformscikit-learn groupbypandas loc and ilocpandas merge and joinpandas transform and filterpandas

29 August 2026

New modules

Seven new modules, all of them in the FastAPI track.

Background TasksFastAPI Lifespan EventsFastAPI OpenAPI and the DocsFastAPI Project StructureFastAPI Security BasicsFastAPI TestingFastAPI async def or defFastAPI

27 August 2026

New modules

19 new modules, all of them in the FastAPI track.

APIRouterFastAPI Class DependenciesFastAPI Dependencies with yieldFastAPI Dependency InjectionFastAPI Dependency OverridesFastAPI Error HandlingFastAPI Form Data and File UploadsFastAPI HTTP Methods and RoutingFastAPI Headers and CookiesFastAPI Path ParametersFastAPI Query ParametersFastAPI Reading a 422FastAPI Request BodiesFastAPI Response ModelsFastAPI Router and Global DependenciesFastAPI Status CodesFastAPI Sub-dependenciesFastAPI What FastAPI IsFastAPI Your First EndpointFastAPI

26 August 2026

New modules

30 new modules, all of them in the Pydantic track.

25 August 2026

New modules

30 new modules: 20 in Maths for Machine Learning and ten in Deep Learning.

24 August 2026

Sixty modules across Computer Vision, Databases and machine learning

Computer Vision gained the classical floor it never had - convolution kernels you edit by hand, thresholding, histograms, the Gaussian/median/bilateral comparison, morphology, colour spaces, interpolation and affine transforms - and then the modern half: separable and dilated convolutions, anchor boxes, mAP, template matching, Harris corners, ViT patching, Grad-CAM and the semantic/instance/panoptic distinction.
The database track picked up schema design and concurrency - keys, constraints, EXISTS against IN against JOIN, recursive CTEs, query plans, covering indexes, isolation levels and deadlocks - then scale and safety: MVCC, partitioning, sharding, replication lag, CAP, the document and graph models, columnar storage, star schemas and parameterised queries.
Machine learning gained the workflow around the algorithms it already had: scaling, leakage, missing values, outliers, pipelines, the precision/recall family, threshold tuning and learning curves, then DBSCAN, hierarchical clustering, choosing k, Gaussian mixtures, embeddings, permutation importance, SHAP, calibration, isolation forests and ensembles.
Every number on those pages is computed in your browser from data generated in your browser. The leakage module really does fit a scaler on the whole dataset and really does report the inflated score that follows; the calibration slider really does leave AUC untouched while wrecking the probabilities.

20 new modules: ten in Databases & SQL and ten in Machine Learning.

Choosing k: Elbow and SilhouetteMachine Learning DBSCAN: Density-Based ClusteringMachine Learning Feature and Permutation ImportanceMachine Learning Gaussian Mixture ModelsMachine Learning Hierarchical Clustering and DendrogramsMachine Learning Isolation ForestMachine Learning Key-Value and Graph ModelsDatabases & SQL MVCC: How Readers Avoid Blocking WritersDatabases & SQL OLTP, OLAP and Columnar StorageDatabases & SQL Partial Dependence, ICE and SHAPMachine Learning Partitioning by Range and HashDatabases & SQL Probability CalibrationMachine Learning Replication, Read Replicas and LagDatabases & SQL SQL Injection and Parameterised QueriesDatabases & SQL ShardingDatabases & SQL Stacking and Voting EnsemblesMachine Learning Star Schema: Facts and DimensionsDatabases & SQL The CAP TheoremDatabases & SQL The Document Model against RowsDatabases & SQL t-SNE and UMAP beside PCAMachine Learning

22 August 2026

New modules

Ten new modules, all of them in the Computer Vision track.

Anchor BoxesComputer Vision Depthwise Separable ConvolutionComputer Vision Dilated ConvolutionsComputer Vision Global Average Pooling against FlattenComputer Vision Grad-CAMComputer Vision Harris Corners and KeypointsComputer Vision Mean Average PrecisionComputer Vision Semantic, Instance and Panoptic SegmentationComputer Vision Template MatchingComputer Vision Vision Transformer PatchesComputer Vision

21 August 2026

Longer articles everywhere, and the labs documented

Every article on the site now runs to full length. The interview track had been quietly losing its written answers on every build - the generator ran after the article writer and overwrote it - which is fixed, and the 34 generated Python modules were expanded to match the hand-written ones.
The four labs - Python, SQL, JavaScript and HTML - now explain how each runtime works, what it supports, the errors you will actually hit and what is worth trying, rather than being an editor with a sentence above it.
Article prose is set in a proportional face at a comfortable reading measure, code inside articles sits in real code blocks, and the column widths were rebalanced so the text is not swimming in space.

30 new modules: ten in Computer Vision, ten in Databases & SQL and ten in Machine Learning.

1x1 ConvolutionsComputer Vision Affine TransformsComputer Vision Aggregate Functions and the NULL TrapDatabases & SQL Blur: Gaussian, Median and BilateralComputer Vision Colour Spaces: RGB and HSVComputer Vision Composite and Covering IndexesDatabases & SQL Constraints: UNIQUE, CHECK and NOT NULLDatabases & SQL Convolution Kernels by HandComputer Vision Data LeakageMachine Learning DeadlocksDatabases & SQL EXISTS, IN and JOINDatabases & SQL EXPLAIN and Query PlansDatabases & SQL Erosion and DilationComputer Vision Feature ScalingMachine Learning Grid Search against Random SearchMachine Learning Handling Missing ValuesMachine Learning Histograms and EqualisationComputer Vision Isolation LevelsDatabases & SQL Learning CurvesMachine Learning Outliers and InfluenceMachine Learning PipelinesMachine Learning Precision, Recall and F1Machine Learning Precision-Recall against ROCMachine Learning Primary and Foreign KeysDatabases & SQL Receptive FieldComputer Vision Recursive CTEsDatabases & SQL Resizing and InterpolationComputer Vision Self-JoinsDatabases & SQL Threshold TuningMachine Learning ThresholdingComputer Vision

18 August 2026

A new theme, and split-pane labs

The whole site moved to the Ash palette, chosen for long reading sessions in either mode: no yellow cast in light mode, nothing pure-black in dark, and a single accent split into two roles so text and fills can each meet contrast on their own terms.
The four labs became genuine split-pane editors - code on one side, output on the other, with a divider you can drag, nudge with the arrow keys or reset with a double-click. The position is remembered per lab.
The Python track grew from twelve modules to forty-six, covering the things a beginner meets in the first week and the generator gaps above them.

34 new modules, all of them in the Python track.

14 August 2026

An interview track, and runnable code inside the articles

A tenth track at /interview/ answers one interview question per page. Each has a step-through visualiser you drive yourself, the full solution in an editor you can run and edit, and questions at the end.
Algorithm modules now carry a working Python program alongside the animation, annotated line by line, so the thing you just watched and the thing you can run are on the same page.
The retrieval and evaluation articles in Gen AI got interactive panels of their own - chunkers, indexes, rankers and the metrics that score them, each with parameters you can move and a readout that explains what moved.
Reader-facing counts are derived from the catalog now rather than typed in, so the site stops claiming a module total it outgrew.

40 new modules: 20 in Generative AI & LLMs and 20 in Interview Questions.

13 August 2026

New modules

29 new modules, all of them in the Interview Questions track.

11 August 2026

JavaScript and HTML labs

A JavaScript compiler at /js-lab/ runs on your browser's own engine, in a worker thread, with the console mirrored back into the page.
An HTML playground at /html-lab/ renders your markup and inline scripts in a sandboxed preview, with console output shown below it.
Both run entirely offline - no interpreter to download, nothing uploaded - joining the Python compiler and SQL playground.

6 August 2026

New modules

15 new modules: twelve in Python and three in Generative AI & LLMs.

Booleans and ComparisonsPython DictionariesPython For Loops and range()Python Functions and Return ValuesPython Hello, Python!Python If, Elif and ElsePython Lists and IndexingPython Multi-Query RetrieverGenerative AI & LLMs Numbers and OperatorsPython Parent Document RetrieverGenerative AI & LLMs Reading Errors and TracebacksPython Self-Query RetrieverGenerative AI & LLMs Strings and SlicingPython Variables and TypesPython While Loops, break and continuePython

5 August 2026

Bookmarks, notes, a glossary and a print stylesheet

Bookmark any module from its header, with a note if you want one, and read them all back on /saved/. Stored in your browser only.
A glossary of 49 terms at /glossary/, and a definition card on the first mention of each term in any article.
Printing a module now gives you a clean study sheet instead of a near-black page: no header, no rail, no dead widgets, and the check answers unfolded.
Keyboard access finished. 52 modules had no keyboard route at all - the ones driven by buttons, checkboxes or number fields rather than sliders. Now 223 of 225 do.
Fixed: the arrow-key hint was invisible on every canvas-driven module, because it was being appended inside the <canvas>.

Four new modules, all of them in the Computer Vision track.

IoU and Non-Max SuppressionComputer Vision Object Detection with Bounding BoxesComputer Vision ResNet and Identity ShortcutsComputer Vision Semantic Segmentation and U-NetComputer Vision

4 August 2026

New modules

17 new modules: eight in Generative AI & LLMs, six in Databases & SQL and three in Deep Learning.

BM25 and Sparse Lexical RetrievalGenerative AI & LLMs CASE and Views in SQLDatabases & SQL Chunking Strategies for RAGGenerative AI & LLMs Dot Product vs Cosine Similarity for RetrievalGenerative AI & LLMs Gradient ClippingDeep Learning Hybrid Search: Dense + Sparse (Reciprocal Rank Fusion)Generative AI & LLMs Layer NormalizationDeep Learning Maximal Marginal Relevance (MMR)Generative AI & LLMs NULL Handling and COALESCE in SQLDatabases & SQL Normalization (1NF, 2NF, 3NF) in SQLDatabases & SQL Query Execution Order in SQLDatabases & SQL Query Rewriting and HyDEGenerative AI & LLMs Re-ranking: Bi-Encoders vs Cross-EncodersGenerative AI & LLMs Residual and Skip ConnectionsDeep Learning Retrieval Evaluation MetricsGenerative AI & LLMs Transactions and ACID in SQLDatabases & SQL UNION, INTERSECT and EXCEPT in SQLDatabases & SQL

3 August 2026

Practice mode, shareable state, and an installable app

Practice at /practice/ draws questions from the modules you have opened, weighted by what you got wrong and how long ago.
Every module's controls are now in its URL, so you can link someone to exactly the state you are looking at.
VizLearn is installable and works offline.
The visualisations became keyboard-operable.

18 new modules: six in Maths for Machine Learning, five in Generative AI & LLMs, four in Databases & SQL, one in Deep Learning, one in Machine Learning and one in Natural Language Processing.

Backpropagation and the Computational GraphDeep Learning Context Windows and the KV CacheGenerative AI & LLMs Cross-Entropy and KL DivergenceMaths for Machine Learning DeterminantMaths for Machine Learning Embeddings and Vector SearchGenerative AI & LLMs Encoder-only vs Decoder-only (BERT vs GPT)Natural Language Processing Fine-Tuning vs RLHF and DPOGenerative AI & LLMs HAVING in SQLDatabases & SQL Hallucination and GroundingGenerative AI & LLMs Identity, Inverse and TransposeMaths for Machine Learning Indexes and Query PerformanceDatabases & SQL Information GainMaths for Machine Learning ORDER BY in SQLDatabases & SQL Principal Component AnalysisMachine Learning ProjectionsMaths for Machine Learning Rank and Linear IndependenceMaths for Machine Learning Retrieval-Augmented Generation (RAG)Generative AI & LLMs Subqueries in SQLDatabases & SQL

1 August 2026

Topic landing pages and the lab layer

Each track got its own page at /<track>/ instead of only existing as a fragment of the hub.
Every module gained a runnable experiment, a predict-then-reveal step, and a check at the end.
about, contact, privacy and terms.

20 new modules: eight in Maths for Machine Learning, six in Natural Language Processing, five in Machine Learning and one in Deep Learning.

Module additions on this page are read from the repository history rather than kept by hand, so nothing here can drift from what actually shipped.