Shamsuddin Ahmed, senior software engineer in Dhaka: agent harnesses, LLM serving infrastructure, and machine learning for biology. Long-form notes on all three.https://shamspias.com/en-gb© 2026 Shamsuddin Ahmedhttps://shamspias.com/posts/2026/08/kinetix-clinical-agent/https://shamspias.com/posts/2026/08/kinetix-clinical-agent/Sixty tools, twenty-nine of them mutating, inside a system holding injury records. The Go that enforces one rule: the agent inherits the caller.Thu, 20 Aug 2026 00:00:00 GMTagent harnessGoclinical softwarebiomechanicsvalidationmeasurementhttps://shamspias.com/posts/2026/08/an-agent-is-data-not-code/https://shamspias.com/posts/2026/08/an-agent-is-data-not-code/Store agents as versioned, validated rows instead of writing them as classes, and rollback, diffs, audit, teams and lessons come free. Here is the bill.Sat, 08 Aug 2026 00:00:00 GMTagent harnessLLMarchitecturesystem designAI engineeringhttps://shamspias.com/posts/2026/07/what-it-refuses-to-measure/https://shamspias.com/posts/2026/07/what-it-refuses-to-measure/Three measured refusals in a markerless movement pipeline, the gaps that are still open, and why a clinician should prefer software that says not measured.Sat, 25 Jul 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationvalidationmeasurementclinical softwareagent harnesshttps://shamspias.com/posts/2026/07/crop-disease-from-the-sky/https://shamspias.com/posts/2026/07/crop-disease-from-the-sky/A frozen vision backbone, a linear probe trained on forty photos, and a three-tier fallback so that a drone survey works on day one.Sat, 11 Jul 2026 00:00:00 GMTcomputer visionagricultureDINOv2linear probingtransfer learningdeep learninghttps://shamspias.com/posts/2026/06/triangulation-and-cameras/https://shamspias.com/posts/2026/06/triangulation-and-cameras/The multi-view path: DLT in enough detail to implement, how many cameras earn their place, and three lessons about evaluating a solver whose truth you cannot see.Sat, 27 Jun 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationmulti-viewtriangulationcalibrationvalidationhttps://shamspias.com/posts/2026/06/keypoints-to-clinical-metrics/https://shamspias.com/posts/2026/06/keypoints-to-clinical-metrics/Two calibrated cameras, DLT triangulation, trajectory filtering, and honest uncertainty intervals: what turns a pose demo into a record a clinician will sign.Sat, 13 Jun 2026 00:00:00 GMTbiomechanicscomputer visionRTMPosesports sciencesystem designmulti-viewhttps://shamspias.com/posts/2026/05/camera-grades-its-own-footage/https://shamspias.com/posts/2026/05/camera-grades-its-own-footage/Camera obliquity is invisible to one lens, so the system infers it from anatomy. Banding its own error by that estimate beats banding by ground truth.Sat, 23 May 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationRTMPosevalidationmeasurementsports sciencehttps://shamspias.com/posts/2026/05/bowling-biomechanics-pose/https://shamspias.com/posts/2026/05/bowling-biomechanics-pose/Joint angles from pose estimation, event detection inside a 100-millisecond delivery swing, and the frame rate that decides what you are allowed to claim.Sat, 02 May 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationMediaPipesports sciencecrickethttps://shamspias.com/posts/2026/04/facing-and-sign/https://shamspias.com/posts/2026/04/facing-and-sign/Which way an athlete faces sets the sign of every sagittal angle at once. One bit, 100 out of 100 on 919 comparisons, and why sign errors are not small.Sat, 25 Apr 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationRTMPosevalidationmeasurementsports sciencehttps://shamspias.com/posts/2026/04/docking-without-jargon/https://shamspias.com/posts/2026/04/docking-without-jargon/What a docking score really is, why the pose matters more than the number, six ways docking lies to you, and where 2026 co-folding models fit.Sat, 04 Apr 2026 00:00:00 GMTdrug discoverymolecular dockingcheminformaticsstructural biologybioinformaticsbeginnerhttps://shamspias.com/posts/2026/03/why-2d-beat-3d/https://shamspias.com/posts/2026/03/why-2d-beat-3d/Lifting one camera to 3D sounds strictly better than measuring in the image plane. On sagittal flexion it is worse, and the reason is a free parameter.Sat, 28 Mar 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationmeasurementvalidationsports sciencemulti-viewhttps://shamspias.com/posts/2026/03/screening-400k-natural-products/https://shamspias.com/posts/2026/03/screening-400k-natural-products/A funnel from every catalogued natural product down to twenty compounds worth testing, and why the split and the base rate matter more than the model.Sat, 14 Mar 2026 00:00:00 GMTdrug discoveryvirtual screeningCOCONUTcheminformaticsmachine learningbioinformaticshttps://shamspias.com/posts/2026/02/pixels-to-joint-angles/https://shamspias.com/posts/2026/02/pixels-to-joint-angles/The whole monocular pipeline for one number, and what the detector's pixel error costs once you express it in degrees instead of in average precision.Sat, 28 Feb 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationRTMPosemeasurementvalidationsports sciencehttps://shamspias.com/posts/2026/02/honest-negatives-peptide-benchmark/https://shamspias.com/posts/2026/02/honest-negatives-peptide-benchmark/Random UniProt negatives quietly turn peptide prediction into short versus long. Here is what an honest benchmark costs, and why it is worth paying.Sat, 14 Feb 2026 00:00:00 GMTbioinformaticspeptidesbenchmarkingmachine learningresearch integritydataset designhttps://shamspias.com/posts/2026/01/joint-angle-accuracy/https://shamspias.com/posts/2026/01/joint-angle-accuracy/A percentage without a tolerance is not a claim. The full scorecard for a video biomechanics system, by configuration, and why 10 degrees settles it.Sat, 31 Jan 2026 00:00:00 GMTbiomechanicscomputer visionpose estimationRTMPosesports sciencevalidationmeasurementhttps://shamspias.com/posts/2026/01/protein-language-models/https://shamspias.com/posts/2026/01/protein-language-models/A 650-million-parameter protein language model, what its embeddings really encode, and why twenty lines of letter counting still beat it on 13-residue peptides.Sat, 17 Jan 2026 00:00:00 GMTbioinformaticsprotein language modelsESM-2embeddingstransfer learningmachine learninghttps://shamspias.com/posts/2025/12/safe-by-default-agents/https://shamspias.com/posts/2025/12/safe-by-default-agents/Read/write enforcement, principal scoping, approval gates, sandboxes and budgets: five safety properties, every one enforced in code, not in a prompt.Sat, 06 Dec 2025 00:00:00 GMTagent harnessLLMsecurityprompt injectionAI engineeringhuman in the loophttps://shamspias.com/posts/2025/11/llm-security-best-practices/https://shamspias.com/posts/2025/11/llm-security-best-practices/The whole series as a checklist, plus the one habit that generates it: assume the model is compromised, and design so that assumption is survivable.Thu, 20 Nov 2025 00:00:00 GMTsecurityllm securityagentsbest practiceshttps://shamspias.com/posts/2025/11/peptide-feature-engineering/https://shamspias.com/posts/2025/11/peptide-feature-engineering/AAC, DPC, DDE, CKSAAP, CTD, PAAC and QSO: every way I know to turn a peptide into a fixed-length vector, and what each one throws away.Sat, 08 Nov 2025 00:00:00 GMTbioinformaticspeptidesfeature engineeringmachine learningcomputational biologyhttps://shamspias.com/posts/2025/10/verbs-not-tables/https://shamspias.com/posts/2025/10/verbs-not-tables/Why intent-named operations beat database schemas, auto-generated OpenAPI specs and one-command MCP servers when you hand an application to a model.Sat, 11 Oct 2025 00:00:00 GMTagent harnessLLMtool usetext-to-SQLapi designAI engineeringhttps://shamspias.com/posts/2025/10/exfiltration-through-model-output/https://shamspias.com/posts/2025/10/exfiltration-through-model-output/The model's output came from your trusted system, so it feels safe. It is not: untrusted input shaped it, and one rendered image can carry your data to the attacker.Thu, 02 Oct 2025 00:00:00 GMTsecurityllm securityexfiltrationprompt injectionhttps://shamspias.com/posts/2025/09/peptides-101/https://shamspias.com/posts/2025/09/peptides-101/What a peptide actually is, why order and side chains carry the information, and why 13 residues is a different modelling problem from 300.Sun, 21 Sep 2025 00:00:00 GMTbioinformaticspeptidescomputational biologymachine learningdrug discoverybeginnerhttps://shamspias.com/posts/2025/09/sensitive-data-pii-secrets/https://shamspias.com/posts/2025/09/sensitive-data-pii-secrets/The safest data is the data the model never sees. Masking PII, keeping secrets out of context, and stopping leaks into logs, training, and other users.Thu, 04 Sep 2025 00:00:00 GMTsecurityllm securitypiiprivacyhttps://shamspias.com/posts/2025/08/what-is-an-agent-harness/https://shamspias.com/posts/2025/08/what-is-an-agent-harness/The five parts of an agent harness: capability surface, meaning, policy, approval and audit. Only two of them are about power.Sat, 30 Aug 2025 00:00:00 GMTagent harnessLLMAI engineeringtool usearchitecturebeginnerhttps://shamspias.com/posts/2025/08/agent-blast-radius/https://shamspias.com/posts/2025/08/agent-blast-radius/A chatbot that is wrong is embarrassing. An agent that is wrong took an action. The security of a tool-calling agent is the security of what you connected it to.Wed, 13 Aug 2025 00:00:00 GMTsecurityllm securityagentsagent harnesshttps://shamspias.com/posts/2025/07/slowest-kid-moe-straggler/https://shamspias.com/posts/2025/07/slowest-kid-moe-straggler/Why a mixture-of-experts layer runs at the speed of its busiest expert, how to measure it, and when predictive prefetching actually helps.Fri, 18 Jul 2025 00:00:00 GMTmixture of expertsneural networksperformancestraggler problemAImachine learningsimple explanationshttps://shamspias.com/posts/2025/07/prompt-injection/https://shamspias.com/posts/2025/07/prompt-injection/The signature attack on language models. How instructions hide in the text a model reads, why you cannot filter your way out, and what containment looks like.Wed, 02 Jul 2025 00:00:00 GMTsecurityllm securityprompt injectionagentshttps://shamspias.com/posts/2025/06/why-ai-security-is-different/https://shamspias.com/posts/2025/06/why-ai-security-is-different/Every security bug in the last series was data mistaken for code. A language model makes that its whole way of working: to it, all text is instructions.Wed, 11 Jun 2025 00:00:00 GMTsecurityllm securityprompt injectionagentshttps://shamspias.com/posts/2025/05/dify-ssl-zero-to-green/https://shamspias.com/posts/2025/05/dify-ssl-zero-to-green/Let's Encrypt certificates on a self-hosted Dify behind Nginx: the ACME flow, the nginx reload everyone forgets, and how to test without burning rate limits.Thu, 08 May 2025 00:00:00 GMTdifySSLLet's EncryptDockerDevOpsself-hostingbeginnerhttps://shamspias.com/posts/2025/05/rag-retrieval-benchmark/https://shamspias.com/posts/2025/05/rag-retrieval-benchmark/How to benchmark sparse, dense, hybrid and late-interaction retrieval on your own corpus, and how to read the numbers without fooling yourself.Thu, 01 May 2025 00:00:00 GMTretrieval-augmented generationsimilarity searchdense vectorsBM25hybrid retrievalevaluationbenchmarksAI foundationshttps://shamspias.com/posts/2025/04/lexsublm-lite/https://shamspias.com/posts/2025/04/lexsublm-lite/A laptop-sized toolkit for context-aware lexical substitution, and an honest look at what tiny models actually score on it.Thu, 24 Apr 2025 00:00:00 GMTNLPopen-sourcemachine learningpythonlexical substitutionhttps://shamspias.com/posts/2025/02/moe-explained-simply/https://shamspias.com/posts/2025/02/moe-explained-simply/How sparse mixture-of-experts really works: per-token routing, top-k gating, load balancing, and the memory bill you pay for the compute you save.Sat, 08 Feb 2025 00:00:00 GMTmixture of expertsneural networksAI basicsmachine learningsimple explanationseducationalMoE fundamentalshttps://shamspias.com/posts/2024/10/network-security-and-tls/https://shamspias.com/posts/2024/10/network-security-and-tls/On a network the confusion is not data and code but identity. What HTTPS really guarantees, what it does not, and why a padlock is about the pipe, not the site.Tue, 15 Oct 2024 00:00:00 GMTsecuritynetwork securitytlscryptographyhttps://shamspias.com/posts/2024/09/binary-exploitation-basics/https://shamspias.com/posts/2024/09/binary-exploitation-basics/The oldest confusion of data and code, at the lowest level. How writing past a buffer can hijack a program, and the four defences that made it hard.Tue, 10 Sep 2024 00:00:00 GMTsecuritybinary exploitationmemory safetychttps://shamspias.com/posts/2024/08/retrieval-metrics/https://shamspias.com/posts/2024/08/retrieval-metrics/What BM25, Recall@k, answer-in-context rate and token F1 actually measure, which number to fix first, and how to compute them without fooling yourself.Thu, 15 Aug 2024 00:00:00 GMTinformation retrievalRAGevaluationBM25dense retrievalquestion answeringmetricsbeginnerhttps://shamspias.com/posts/2024/07/broken-access-control/https://shamspias.com/posts/2024/07/broken-access-control/The most common serious web bug is also the most boring: the server checks who you are, then forgets to check whether you may touch this particular thing.Tue, 09 Jul 2024 00:00:00 GMTsecurityweb securityaccess controlauthorizationhttps://shamspias.com/posts/2024/06/cross-site-request-forgery/https://shamspias.com/posts/2024/06/cross-site-request-forgery/The attacker never steals the session. They get the victim's own browser to send an authenticated request, using cookies the browser attaches for them.Tue, 11 Jun 2024 00:00:00 GMTsecurityweb securitycsrfcookieshttps://shamspias.com/posts/2024/05/cross-site-scripting/https://shamspias.com/posts/2024/05/cross-site-scripting/The same bug as SQL injection, moved to the browser. How attacker text becomes attacker script in your users' sessions, the three flavours, and the layered fix.Tue, 14 May 2024 00:00:00 GMTsecurityweb securityxssbrowsershttps://shamspias.com/posts/2024/04/sql-injection/https://shamspias.com/posts/2024/04/sql-injection/The clearest case of data becoming code. How a login form ends up running the attacker's SQL, why escaping is the wrong fix, and the one that actually ends it.Tue, 09 Apr 2024 00:00:00 GMTsecurityweb securitysql injectiondatabaseshttps://shamspias.com/posts/2024/03/what-security-actually-is/https://shamspias.com/posts/2024/03/what-security-actually-is/Security is not a product you add. It is a property of where you decide to trust, and almost every vulnerability is one of two confusions crossing that line.Tue, 12 Mar 2024 00:00:00 GMTsecurityweb securitythreat modellingtrust boundarieshttps://shamspias.com/posts/2024/02/classical-planning/https://shamspias.com/posts/2024/02/classical-planning/PDDL, heuristics that sometimes do nothing, HTN, temporal and multi-agent planning, with real planner numbers on Towers of Hanoi.Fri, 09 Feb 2024 00:00:00 GMTclassical planningsearchPDDLHTNmulti-agent systemsknowledge representationAI foundationsbeginnerhttps://shamspias.com/posts/2024/01/classical-planning/https://shamspias.com/posts/2024/01/classical-planning/How STRIPS turns a to-do list into an algorithm: factored states, progression and regression, delete relaxation, planning graphs, and modern planners.Mon, 15 Jan 2024 00:00:00 GMTartificial intelligenceplanningstripsstate-space searchplanning graphhierarchical planningbeginnerhttps://shamspias.com/posts/2023/12/first-order-logic/https://shamspias.com/posts/2023/12/first-order-logic/Quantifiers, unification, forward and backward chaining, and resolution, from Socrates to a solver that finds the proof for you.Mon, 25 Dec 2023 00:00:00 GMTartificial intelligencefirst-order logicunificationforward chainingbackward chainingresolutionbeginnerhttps://shamspias.com/posts/2023/12/logical-agents/https://shamspias.com/posts/2023/12/logical-agents/Knowledge bases, entailment, resolution and SAT solving, learned by staying alive in the Wumpus World, with runnable PySAT code.Sat, 02 Dec 2023 00:00:00 GMTartificial intelligenceknowledge representationpropositional logictheorem provingsat solverwumpus worldbeginnerhttps://shamspias.com/posts/2023/09/csp-basics/https://shamspias.com/posts/2023/09/csp-basics/Arc consistency, backtracking with MRV and min-conflicts, explained through Sudoku and map colouring, plus what you would actually use today.Sat, 16 Sep 2023 00:00:00 GMTartificial intelligenceconstraint satisfactioncspbacktrackingarc consistencysudokubeginnerhttps://shamspias.com/posts/2023/08/adversarial-search/https://shamspias.com/posts/2023/08/adversarial-search/Minimax, alpha-beta pruning, expectiminimax and MCTS: how a program plans when something is planning against it, and what neural evaluation changed.Sat, 05 Aug 2023 00:00:00 GMTartificial intelligencegame aiadversarial searchminimaxalpha-beta pruningmonte carlo tree searchbeginnerhttps://shamspias.com/posts/2023/07/beyond-search/https://shamspias.com/posts/2023/07/beyond-search/Hill climbing, simulated annealing, belief states and online search: what to do when A* is the wrong tool, the world hides half of itself, or there is no map yet.Wed, 26 Jul 2023 00:00:00 GMTartificial intelligencesearch algorithmslocal searchstochasticpartial observabilityonline learningbeginnerhttps://shamspias.com/posts/2023/05/backend-best-practices/https://shamspias.com/posts/2023/05/backend-best-practices/Boring is the highest compliment a backend can earn. The layering, config, jobs, and observability habits that stop the phone call at three in the morning.Tue, 16 May 2023 00:00:00 GMTbackendbest practicessoftware engineeringarchitecturehttps://shamspias.com/posts/2023/04/ai-agents-search/https://shamspias.com/posts/2023/04/ai-agents-search/What an agent actually is, how to turn a problem into a search problem, and why changing one line of priority code turns breadth-first search into A*.Wed, 26 Apr 2023 00:00:00 GMTartificial intelligenceintelligent agentssearch algorithmsheuristicbeginnerpythonhttps://shamspias.com/posts/2023/02/ml-101-pytorch-tf-decision-tree/https://shamspias.com/posts/2023/02/ml-101-pytorch-tf-decision-tree/What a deep-learning framework actually does for you, how a decision tree picks its questions, and which of the three to reach for on real data.Mon, 13 Feb 2023 00:00:00 GMTmachine learningdeep learningpytorchtensorflowdecision treebeginnerpythonhttps://shamspias.com/posts/2023/01/math-for-ai-linear-algebra-basics/https://shamspias.com/posts/2023/01/math-for-ai-linear-algebra-basics/Scalars, vectors, matrices and tensors, the five operations every neural network is made of, and the shape rules that cause most of the bugs.Mon, 23 Jan 2023 00:00:00 GMTmathAIlinear algebrabeginnerpythonhttps://shamspias.com/posts/2022/10/grpc-between-services/https://shamspias.com/posts/2022/10/grpc-between-services/REST is built for the browser. When both ends are your own servers, a typed binary contract with generated code is often faster and safer. That is gRPC.Tue, 11 Oct 2022 00:00:00 GMTbackendgrpcmicroservicesprotobufhttps://shamspias.com/posts/2022/06/transformers-attention-made-simple/https://shamspias.com/posts/2022/06/transformers-attention-made-simple/Attention and the transformer block from first principles, plus what changed by 2026: pre-norm, RMSNorm, RoPE, grouped-query attention and the KV cache.Mon, 06 Jun 2022 00:00:00 GMTAItransformersdeep learningattentionbeginnereducationhttps://shamspias.com/posts/2022/02/real-time-polling-websocket-webrtc/https://shamspias.com/posts/2022/02/real-time-polling-websocket-webrtc/HTTP is a client asking and a server answering. Real-time is the server needing to speak first. Here is the ladder of ways to do that, and what to pick when.Tue, 15 Feb 2022 00:00:00 GMTbackendwebsocketwebrtcnetworkinghttps://shamspias.com/posts/2021/11/memory-and-why-arrays-win/https://shamspias.com/posts/2021/11/memory-and-why-arrays-win/Big-O counts steps and assumes every step costs the same. It does not. A cache miss costs a hundred steps, and that is why a linked list loses to an array.Sat, 20 Nov 2021 00:00:00 GMTalgorithmsmemorycacheperformanceproblem solvinghttps://shamspias.com/posts/2021/07/ml-algorithms-by-hand/https://shamspias.com/posts/2021/07/ml-algorithms-by-hand/Four algorithms in forty lines each, with no library. k-nearest neighbours, linear and logistic regression, k-means, and a decision tree, plus what each one assumes.Sat, 24 Jul 2021 00:00:00 GMTalgorithmsmachine learningregressionk-meansdecision treeshttps://shamspias.com/posts/2021/06/rest-api-that-ages-well/https://shamspias.com/posts/2021/06/rest-api-that-ages-well/An API is a contract you have to keep. Resources and verbs, the status codes that matter, idempotency, pagination, versioning, and errors that a client can act on.Tue, 15 Jun 2021 00:00:00 GMTbackendrest apiapi designhttphttps://shamspias.com/posts/2021/03/fenwick-and-segment-trees/https://shamspias.com/posts/2021/03/fenwick-and-segment-trees/Prefix sums break the moment an element changes. Two structures answer range questions and point updates in logarithmic time, and one of them is twenty lines.Sat, 27 Mar 2021 00:00:00 GMTalgorithmsdata structuresfenwick treesegment treeproblem solvinghttps://shamspias.com/posts/2021/03/homomorphic-encryption-guide/https://shamspias.com/posts/2021/03/homomorphic-encryption-guide/What homomorphic encryption really does, which scheme to pick, what a ciphertext costs in bytes and depth, and where it still loses to a trusted enclave.Sat, 06 Mar 2021 00:00:00 GMTprivacyencryptionmachine learninghomomorphic encryptionAI securityhttps://shamspias.com/posts/2020/11/strings-hashing-kmp-z/https://shamspias.com/posts/2020/11/strings-hashing-kmp-z/Where does this pattern occur? Three answers in linear time, each with a different trade: one is easy and probabilistic, two are exact and need a table.Sat, 14 Nov 2020 00:00:00 GMTalgorithmsstringshashingkmpproblem solvinghttps://shamspias.com/posts/2020/09/sql-or-nosql/https://shamspias.com/posts/2020/09/sql-or-nosql/NoSQL was sold as the future and is really a specialised tool. What each kind buys, why relational is the right default, and the real cases for reaching past it.Tue, 15 Sep 2020 00:00:00 GMTbackenddatabasesnosqlpostgresqlhttps://shamspias.com/posts/2020/07/number-theory-you-need/https://shamspias.com/posts/2020/07/number-theory-you-need/Sieves, greatest common divisors, modular arithmetic and inverses. Five tools, each about ten lines, that cover almost every number theory problem you will meet.Sun, 19 Jul 2020 00:00:00 GMTalgorithmsnumber theorymodular arithmeticprimesproblem solvinghttps://shamspias.com/posts/2020/04/databases-and-acid/https://shamspias.com/posts/2020/04/databases-and-acid/A transaction is a promise the database makes when two things happen at once. What ACID guarantees, the anomalies it prevents, and the race it does not.Tue, 21 Apr 2020 00:00:00 GMTbackenddatabasesacidpostgresqlhttps://shamspias.com/posts/2020/03/game-theory-nim-and-grundy/https://shamspias.com/posts/2020/03/game-theory-nim-and-grundy/Two players, perfect information, no luck. Who wins? The answer comes from labelling positions, and one theorem collapses any such game into a single number.Sun, 08 Mar 2020 00:00:00 GMTalgorithmsgame theorynimgrundyproblem solvinghttps://shamspias.com/posts/2019/11/disjoint-sets-and-mst/https://shamspias.com/posts/2019/11/disjoint-sets-and-mst/Twelve lines of code, two operations, and a running time so close to constant that nobody bothers with the difference. Then minimum spanning trees for free.Sun, 03 Nov 2019 00:00:00 GMTalgorithmsgraphsunion findminimum spanning treeproblem solvinghttps://shamspias.com/posts/2019/10/django-or-fastapi/https://shamspias.com/posts/2019/10/django-or-fastapi/One gives you everything and an opinion; the other gives you speed and a blank page. The choice is about how much you want the framework to decide for you.Tue, 08 Oct 2019 00:00:00 GMTbackenddjangofastapipythonhttps://shamspias.com/posts/2019/06/shortest-paths/https://shamspias.com/posts/2019/06/shortest-paths/Three algorithms, three different jobs. Which one to reach for depends on negative edges, on how many sources you have, and on nothing else.Sun, 09 Jun 2019 00:00:00 GMTalgorithmsgraphsshortest pathsdijkstraproblem solvinghttps://shamspias.com/posts/2019/05/design-patterns-you-actually-use/https://shamspias.com/posts/2019/05/design-patterns-you-actually-use/The pattern books list twenty-three. You reach for about six. Here are those six, the problem each solves, and why naming a pattern is a tool, not a trophy.Tue, 14 May 2019 00:00:00 GMTbackenddesign patternssoftware engineeringpythonhttps://shamspias.com/posts/2019/02/dfs-cycles-topological-order/https://shamspias.com/posts/2019/02/dfs-cycles-topological-order/BFS goes wide, DFS goes deep, and the difference is not stylistic. DFS answers questions about structure that a queue simply cannot see.Sun, 17 Feb 2019 00:00:00 GMTalgorithmsgraphsdfstopological sortproblem solvinghttps://shamspias.com/posts/2018/10/graphs-and-bfs/https://shamspias.com/posts/2018/10/graphs-and-bfs/Half of all algorithm problems are graph problems in disguise. Here is how to represent a graph, and the search that finds shortest paths for free.Sun, 21 Oct 2018 00:00:00 GMTalgorithmsgraphsbfsproblem solvinghttps://shamspias.com/posts/2018/06/knapsack-and-bitmask-dp/https://shamspias.com/posts/2018/06/knapsack-and-bitmask-dp/Knapsack is one recurrence with a dozen disguises. Learn the family, then learn to put a set of items in the state with a bitmask when n is small.Sun, 17 Jun 2018 00:00:00 GMTalgorithmsdynamic programmingknapsackbitmaskproblem solvinghttps://shamspias.com/posts/2018/02/dp-as-a-table/https://shamspias.com/posts/2018/02/dp-as-a-table/Four questions turn a problem into a dynamic program: what is the state, what is the answer for a state, what is the base, and in what order do you fill it.Sun, 25 Feb 2018 00:00:00 GMTalgorithmsdynamic programmingproblem solvingtabulationhttps://shamspias.com/posts/2017/11/memoisation/https://shamspias.com/posts/2017/11/memoisation/The cheapest optimisation in programming: remember what you already worked out. One dictionary turns an exponential recursion into a linear one.Sun, 12 Nov 2017 00:00:00 GMTalgorithmsmemoisationdynamic programmingproblem solvinghttps://shamspias.com/posts/2017/07/recursion-and-backtracking/https://shamspias.com/posts/2017/07/recursion-and-backtracking/Recursion is hard because everyone traces it. The habit that makes it mechanical: write the contract, trust the smaller call, never trace the stack.Sun, 30 Jul 2017 00:00:00 GMTalgorithmsrecursionbacktrackingproblem solvinghttps://shamspias.com/posts/2017/04/greedy-when-it-works/https://shamspias.com/posts/2017/04/greedy-when-it-works/A greedy rule that passes the samples and fails the tests is the most common way to lose a contest. Here is how to tell a correct greedy rule from a plausible one.Sat, 08 Apr 2017 00:00:00 GMTalgorithmsgreedyproblem solvingproofshttps://shamspias.com/posts/2017/01/binary-search-on-the-answer/https://shamspias.com/posts/2017/01/binary-search-on-the-answer/Searching a sorted array is the easy half. The useful half is guessing the answer and asking a monotone yes-or-no question, with no array in sight.Sat, 21 Jan 2017 00:00:00 GMTalgorithmsbinary searchproblem solvingmonotonicityhttps://shamspias.com/posts/2016/12/pointers-and-linked-lists/https://shamspias.com/posts/2016/12/pointers-and-linked-lists/A pointer is the address of a box, not the box. Once that clicks, dynamic memory, references and the linked list all fall out of the same idea. In C and C++.Sun, 18 Dec 2016 00:00:00 GMTalgorithmspointerslinked listscppproblem solvinghttps://shamspias.com/posts/2016/10/sorting-what-to-know/https://shamspias.com/posts/2016/10/sorting-what-to-know/You will almost never write a sort. You will constantly decide what to sort by, and that decision is where the problems are actually solved.Sat, 15 Oct 2016 00:00:00 GMTalgorithmssortingproblem solvingcomparatorshttps://shamspias.com/posts/2016/07/prefix-sums-and-two-pointers/https://shamspias.com/posts/2016/07/prefix-sums-and-two-pointers/Two techniques that turn a quadratic loop into a linear one, on the humblest data structure there is. Range sums in constant time, and the sliding window.Sat, 09 Jul 2016 00:00:00 GMTalgorithmsarraysprefix sumstwo pointersproblem solvinghttps://shamspias.com/posts/2016/04/big-o-without-the-maths/https://shamspias.com/posts/2016/04/big-o-without-the-maths/Big-O is not a speed and not a stopwatch. It is a claim about the shape of a curve. Six shapes, where each one comes from, and the four rules for reading them.Sat, 23 Apr 2016 00:00:00 GMTalgorithmscomplexitybig-oproblem solvinghttps://shamspias.com/posts/2016/02/counting-the-steps/https://shamspias.com/posts/2016/02/counting-the-steps/Before any algorithm makes sense you need one habit: counting how many times a line runs. Here is that habit, built from four small C++ programs and a stopwatch.Sun, 14 Feb 2016 00:00:00 GMTalgorithmscomplexityproblem solvingbeginners