Understand a little.
Then go deeper.
AI, technology and finance, from lasting foundations to modern methods. Explore classical models, today’s AI systems, the infrastructure behind software and the mechanics of money. Each foundation lesson now teaches the idea in steps, explains unfamiliar terms, walks through an example and gives you practice with explained answers. Take your time; understanding matters more than completion.
Beginner
Build your vocabulary.AI, machine learning and models
Artificial intelligence is a broad field of systems that perform tasks associated with intelligence. Machine learning is an approach in which a model learns patterns from examples.
3 min read Explore lesson ↗AITraining, inference and loss
Training adjusts a model’s parameters using examples. Inference uses the trained parameters to produce an output for a new input. Loss measures an aspect of training error.
3 min read Explore lesson ↗TechnologyAPIs, inputs and outputs
An application programming interface, or API, is an agreed way for one piece of software to interact with another. It specifies available operations and the information they accept or return.
3 min read Explore lesson ↗TechnologyHow a web request works
HTTP is a request-and-response protocol. A client asks a server for a resource or operation, and the server responds with a status, headers and sometimes a body.
3 min read Explore lesson ↗FinanceInterest and compound growth
Interest is a cost of borrowing or a return paid on certain savings. Compounding means interest or returns are added to the base used to calculate subsequent growth.
3 min read Explore lesson ↗FinanceInflation and purchasing power
Inflation is an increase in the general price level over time. Purchasing power describes how much a given amount of money can buy.
3 min read Explore lesson ↗AISymbolic AI, rules and expert systems
Symbolic AI represents knowledge with explicit symbols and rules. An expert system applies a knowledge base to a limited class of problems.
3 min read Explore lesson ↗AISupervised learning: labels and targets
Supervised learning uses examples with known target outputs to learn a mapping that can be applied to new inputs.
3 min read Explore lesson ↗AIUnsupervised learning and clustering
Unsupervised learning looks for structure in data without a supplied target label for every example. Clustering groups items under a chosen similarity measure.
3 min read Explore lesson ↗AILinear regression versus logistic regression
Linear regression models a numerical response. Logistic regression transforms a score into a probability for a binary outcome.
3 min read Explore lesson ↗AINeural networks and activation functions
A neural network combines layers of parameterized transformations. Nonlinear activation functions allow it to model relationships beyond a single linear mapping.
3 min read Explore lesson ↗AITokens, context windows and language models
A token is a unit a model processes, often a word fragment. A context window limits how much tokenized material a model can consider in one request.
3 min read Explore lesson ↗TechnologyBits, bytes and text encoding
A bit has two possible values. A byte usually contains eight bits. Encoding defines how information such as text is represented by those values.
3 min read Explore lesson ↗TechnologyAlgorithms and computational complexity
An algorithm is a procedure for solving a problem. Computational complexity describes how its resource use grows as input size increases.
3 min read Explore lesson ↗TechnologyHTML, CSS and JavaScript: three different jobs
HTML expresses document structure, CSS controls presentation and layout, and JavaScript supplies programmable behavior in the browser.
3 min read Explore lesson ↗TechnologyDomain names, DNS and addresses
The Domain Name System maps names to records, such as addresses or aliases, so clients can locate internet services.
3 min read Explore lesson ↗TechnologyOperating systems, processes and memory
An operating system manages hardware resources and provides services to programs. A process is a running program with execution state and managed resources.
3 min read Explore lesson ↗TechnologyVersion control, commits and branches
Version control records changes to project files. A commit captures a recorded state, while a branch names a line of development.
3 min read Explore lesson ↗FinanceCash flow, budgets and net worth
Cash flow measures money moving in and out during a period. Net worth is assets minus liabilities at a point in time.
3 min read Explore lesson ↗FinanceStocks, ownership and market capitalization
A stock represents an ownership interest in a company. Market capitalization is share price multiplied by the relevant number of outstanding shares.
3 min read Explore lesson ↗FinanceETFs, index funds and what they hold
An exchange-traded fund trades on an exchange and holds or tracks a portfolio. An index fund aims to follow a specified index; not every ETF is an index fund.
3 min read Explore lesson ↗FinanceFees and the difference between gross and net returns
Gross return is measured before specified costs. Net return subtracts those costs; the definition should say which expenses and taxes are included.
3 min read Explore lesson ↗FinanceLiquidity, time horizon and forced selling
Liquidity is the ability to transact promptly without an excessive price concession. Time horizon is how long before money is needed for a goal.
3 min read Explore lesson ↗FinanceBalance sheets, income statements and cash flow
A balance sheet describes financial position at a date. An income statement reports performance over a period. A cash-flow statement explains changes in cash.
3 min read Explore lesson ↗AIFraming decisions, uncertainty and what-if analysis
A decision model separates objectives, available actions, assumptions and uncertain outcomes. What-if analysis explores how a specified change in inputs alters a modeled result.
3 min read Explore lesson ↗AIExploratory analysis, data types and descriptive statistics
Exploratory data analysis examines structure, quality and patterns before formal modeling. Descriptive statistics summarize what the observed data contain, not automatically what a wider population does.
3 min read Explore lesson ↗AIVisual storytelling, interactive charts and Power BI
Data visualization encodes measurements in visual form. A useful data story connects a clear question, appropriate comparisons and uncertainty to a defensible decision.
3 min read Explore lesson ↗AIProbability, distributions and expected values
A probability distribution describes possible outcomes and their likelihoods. It models uncertainty under assumptions; it is not a promise about the next observation.
3 min read Explore lesson ↗AIAI project lifecycles, strategy and prioritization
A project is temporary work intended to create a defined result. Managing an AI project connects that result to a business objective, a delivery approach and evidence of benefit.
3 min read Explore lesson ↗Intermediate
Connect the ideas.Embeddings and retrieval
An embedding represents an item as a vector of numbers. A retrieval system can compare these vectors to find items that are similar under a learned representation.
2 min read Explore lesson ↗AIPrecision, recall and evaluation
Precision asks how many positive predictions were correct. Recall asks how many actual positives were found. These metrics answer different questions about errors.
2 min read Explore lesson ↗TechnologyDatabases and atomic transactions
A transaction groups database operations into a unit that can commit together or roll back together. Atomicity prevents a partial group of changes from being treated as complete.
2 min read Explore lesson ↗TechnologyReliability, latency and SLOs
A service-level indicator measures an aspect of service behavior. A service-level objective, or SLO, sets a target for that indicator over a defined window.
2 min read Explore lesson ↗FinanceBonds, prices and yields
A bond is a debt security with contractual payment terms. Its market price can differ from its face value, and its yield depends on price and expected cash flows.
2 min read Explore lesson ↗FinanceDiversification, correlation and risk
Diversification spreads exposure across investments. Correlation describes how returns move together; owning many items does not necessarily create independent sources of risk.
2 min read Explore lesson ↗AIOverfitting, generalization and regularization
Overfitting occurs when a model fits training-specific patterns that do not transfer well to new data. Regularization constrains learning to improve generalization.
2 min read Explore lesson ↗AIData leakage and honest evaluation
Data leakage occurs when training or evaluation uses information that would not legitimately be available at prediction time.
2 min read Explore lesson ↗AIDecision trees and ensembles
A decision tree makes predictions through successive feature-based splits. Ensembles combine multiple models, often to reduce variance or improve predictive performance.
2 min read Explore lesson ↗AIRetrieval-augmented generation: evidence before answers
Retrieval-augmented generation, or RAG, supplies retrieved external material to a generative model so it can answer using that material.
2 min read Explore lesson ↗AIHallucination, uncertainty and calibration
A hallucination is a generated claim that is unsupported or incorrect. Calibration concerns whether predicted confidence matches observed correctness over many comparable cases.
2 min read Explore lesson ↗AIFine-tuning, adapters and LoRA
Fine-tuning adapts a pretrained model using additional training. LoRA learns low-rank updates to selected weight matrices instead of training every original parameter.
2 min read Explore lesson ↗TechnologyDatabase indexes and query plans
A database index is an additional structure that helps locate rows without scanning the entire table for every query.
2 min read Explore lesson ↗TechnologyCaching, freshness and invalidation
A cache stores a reusable result closer to where it is needed. Invalidation decides when that stored result is no longer acceptable.
2 min read Explore lesson ↗TechnologyAuthentication versus authorization
Authentication establishes who or what is making a request. Authorization determines which actions that identity is allowed to perform.
2 min read Explore lesson ↗TechnologyUnit, integration and end-to-end tests
Testing checks behavior at different boundaries. Unit tests isolate a small part; integration tests check interacting components; end-to-end tests exercise a complete user flow.
2 min read Explore lesson ↗TechnologyContainers, virtual machines and isolation
A container packages an application with dependencies while typically sharing the host kernel. A virtual machine runs a guest operating system over virtualized hardware.
2 min read Explore lesson ↗TechnologyMessage queues, delivery and consumers
A message queue decouples producers from consumers by storing work until it can be processed.
2 min read Explore lesson ↗FinanceMonetary policy and interest-rate transmission
Monetary policy uses central-bank instruments to influence financial conditions and support the central bank’s mandate. Its effects reach households and firms through several channels.
2 min read Explore lesson ↗FinanceYield curves and the term structure
A yield curve plots comparable debt yields against maturity. Its shape reflects several influences, including expected short-term rates and term premiums.
2 min read Explore lesson ↗FinanceCredit risk, default and recovery
Credit risk is the risk that a borrower will not meet promised obligations. Loss depends on both the chance of default and the amount recovered if default occurs.
2 min read Explore lesson ↗FinanceMarket orders, limit orders and spreads
A market order prioritizes execution at available prices. A limit order sets a price boundary but may not execute. The bid-ask spread is the gap between quoted buying and selling prices.
2 min read Explore lesson ↗FinanceExchange rates and foreign investment returns
An exchange rate gives one currency’s value in another. A foreign investment’s home-currency return combines the asset’s return with the currency movement.
2 min read Explore lesson ↗FinanceRebalancing and portfolio drift
Rebalancing brings portfolio weights back toward a chosen allocation after market moves or cash flows change them.
2 min read Explore lesson ↗AIMonte Carlo simulation, dependence and tornado charts
Monte Carlo simulation repeatedly samples uncertain inputs and computes outcomes. Sensitivity analysis examines which assumptions drive results and how conclusions change.
3 min read Explore lesson ↗AIDecision trees, expected monetary value and break-even sensitivity
A decision-analysis tree represents choices, chance events and consequences in their information order. It differs from a machine-learning decision tree used for prediction.
3 min read Explore lesson ↗AIMulti-criteria decisions: direct rating, weights and pairwise comparison
Multi-criteria decision analysis structures choices involving several objectives. Direct ratings score alternatives on criteria; pairwise approaches compare elements two at a time.
3 min read Explore lesson ↗AIFairness, hiring-system bias and social sorting
An algorithmic fairness audit examines who benefits, who is harmed and whether a system’s data, objectives and workflow create unjustified differences or exclusion.
3 min read Explore lesson ↗AIAI ethics frameworks, governance architecture and sandboxes
AI governance assigns decision rights, evidence requirements, controls and accountability across a system’s lifecycle. Ethical principles become useful when they change operational choices.
3 min read Explore lesson ↗AIReflective cognition, AI entrepreneurship and professional judgment
Reflective cognition examines how conclusions were reached, what assumptions shaped them and how alternative evidence might change them. AI can support this process or displace it through uncritical reliance.
3 min read Explore lesson ↗AITurning business problems into mathematical and nonlinear models
A mathematical model represents a business question using defined quantities, relationships, objectives and constraints. Nonlinear relationships allow effects that do not follow one constant slope.
3 min read Explore lesson ↗AITime-series forecasting, seasonality and backtesting
A time series records observations in order. Forecasting estimates future values using information available at a particular time, with uncertainty that depends on horizon and assumptions.
3 min read Explore lesson ↗TechnologyDeploying predictive analytics with Snowflake and operational controls
Deployment makes a model usable in a live workflow. A cloud data warehouse such as Snowflake can support data preparation, scheduled scoring and analytics, but requires explicit permissions and operational ownership.
3 min read Explore lesson ↗AIPredictive versus generative AI, business value and decision strategy
Predictive analytics estimates outcomes or categories; generative AI produces content such as text or images. A useful strategy ties either capability to a measurable workflow and appropriate controls.
3 min read Explore lesson ↗AISamples, sampling distributions and confidence intervals
A sample is a subset used to learn about a population. A sampling distribution describes how a statistic varies across repeated samples under a specified design.
3 min read Explore lesson ↗AIHypothesis testing, A/B experiments and practical significance
A hypothesis test evaluates how compatible observed data are with a specified null model. An A/B experiment assigns alternatives to comparable units to estimate their effects.
3 min read Explore lesson ↗AIData wrangling, variable selection and feature engineering
Data wrangling makes records usable and consistent. Feature engineering creates model inputs, while variable selection chooses which inputs to retain.
3 min read Explore lesson ↗AIStakeholders, organizational structures and project leadership
Stakeholder management identifies affected people, their interests and their decision rights. Organizational structure and culture influence how project authority and communication work.
3 min read Explore lesson ↗AIScope, work breakdown structures and communication plans
Scope defines the work and results included in a project. A work breakdown structure decomposes that scope into manageable deliverables and work packages.
3 min read Explore lesson ↗AIEstimates, critical paths and resource scheduling
A project schedule combines activities, dependencies, duration estimates and resource availability. The critical path constrains the earliest finish under the stated schedule assumptions.
3 min read Explore lesson ↗AICPMAI, AI application patterns and iterative delivery
A data-centered AI project method connects business understanding, data work, modeling, evaluation and operational use through repeated evidence-based decisions.
3 min read Explore lesson ↗Technology / NEW ADDITIONSecrets management: hardcoded credentials, scope and rotation
Learning objective: explain why credentials embedded in software are unsafe, design a safer retrieval path, and respond correctly when a secret is exposed. A secret is a value that grants access, such as a database password, API key or signing key. Secrets management is the controlled process for creating, storing, granting, auditing, rotating, revoking and retiring those values.
4 min read Explore lesson ↗Technology / NEW ADDITIONOffline-first systems: queues, synchronization and conflict resolution
Learning objective: explain how an application can keep working during unreliable connectivity, safely synchronize later, and resolve conflicting updates without silently losing data. An offline-first system treats the local device as a usable working environment rather than a broken copy of the server. It stores necessary data locally, records user actions while disconnected, and exchanges changes with a central service when a connection returns. A conflict occurs when two actors update the same logical record from different versions.
4 min read Explore lesson ↗Finance / NEW ADDITIONCrypto-asset custody: keys, wallets and counterparty risk
Learning objective: explain how control of cryptographic keys determines access to a crypto asset, compare hot and cold wallets, and separate self-custody risk from third-party custodian risk. Custody means the arrangements used to store and access an asset. A wallet usually stores or controls the keys used to authorize transactions; it does not contain the asset itself. A private key authorizes transfers, while a related public key can help others verify transactions or send assets to the corresponding address.
4 min read Explore lesson ↗Technology / NEW ADDITIONData-centre water use: withdrawal, consumption and WUE
Learning objective: distinguish water withdrawal from water consumption, calculate Water Usage Effectiveness, and explain why a data centre’s water impact depends on location, electricity and total workload—not one headline number. Water withdrawal is water taken from a river, aquifer or utility. Water consumption is the portion not returned promptly to the same local system, often because it evaporates. Water Usage Effectiveness, or WUE, expresses water consumed per unit of IT-equipment energy.
4 min read Explore lesson ↗Technology / NEW ADDITIONSubmarine cables: how the internet crosses oceans and survives cuts
Learning objective: explain how submarine fibre-optic cables carry international internet traffic, how landing stations connect them to inland networks, and how redundancy, route diversity and spare capacity reduce the impact of a failure. A landing station is the coastal facility where an undersea cable connects to terrestrial, or land-based, networks. Redundancy means keeping backup components or paths. Route diversity means those alternatives do not share the same physical point of failure.
4 min read Explore lesson ↗AI / NEW ADDITIONAI inference performance: latency, throughput and batching
Learning objective: understand how an AI service trades response speed against total processing capacity, and choose measurements that match the user’s need. Inference is the stage in which a trained model produces an output. Latency is the time one request waits and runs. Throughput is the amount of work completed per unit of time, such as requests or generated tokens per second. Concurrency is the number of requests being processed at once. Batching groups requests so the hardware can process them together.
4 min read Explore lesson ↗Technology / NEW ADDITIONLEO satellite internet: orbits, gateways and moving routes
Learning objective: explain how a low-Earth-orbit satellite network carries an internet packet, why it can reduce propagation delay compared with geostationary satellites, and why moving satellites make routing difficult. Low Earth orbit, or LEO, means an orbit no more than 2,000 kilometres above Earth. A constellation is a coordinated set of satellites. A gateway is a ground station that connects the constellation to terrestrial networks.
4 min read Explore lesson ↗AI / NEW ADDITIONAI Evaluations: Test Sets, Graders, and Regression Checks
Learning objective: design a small, repeatable evaluation that shows whether an AI system is good enough for a specific real-world use. An evaluation, or eval, is a structured test made of representative inputs, expected qualities, a scoring method, and an acceptance threshold. A grader is the rule, program, model, or human judgment used to score each result. A regression is a harmful performance drop after a change.
4 min read Explore lesson ↗Finance / NEW ADDITIONMortgage term versus amortization: read the payment schedule
Learning objective: understand how a mortgage term, amortization period, interest rate and payment schedule fit together, then calculate how one payment is split. Principal is the amount still owed. Interest is the lender’s charge for using the money. The term is the period during which the current contract and rate apply. Amortization is the estimated total time needed to repay the loan if the schedule continues.
4 min read Explore lesson ↗Technology / NEW ADDITIONThreat modeling connected devices: trace data before attackers do
Learning objective: build a simple threat model for a connected device and turn it into testable security requirements. A threat model is a structured description of what a system protects, how data moves, where trust changes, what could go wrong, and how the design reduces risk. An asset is something valuable, such as a voice transcript or account token. A trust boundary is a point where data moves between components with different levels of control.
4 min read Explore lesson ↗Technology / NEW ADDITIONSoftware vendor lock-in: measure the exit before you enter
Learning objective: understand vendor lock-in, estimate the real cost of switching, and design an exit plan before choosing a software or cloud service. Vendor lock-in means that changing suppliers becomes unusually difficult or expensive because your data, code, skills, integrations, contracts, or operating procedures depend on one provider. Portability is the ability to move data or workloads. Interoperability is the ability of different systems to work together.
4 min read Explore lesson ↗AI / NEW ADDITIONSim-to-real learning: the reality gap and domain randomization
Learning objective: explain why an AI system trained in a simulator can fail on hardware, and how domain randomization can make transfer more robust. Sim-to-real transfer means training a model or control policy in a simulated environment, then deploying it in the real world. The reality gap is the mismatch between simulated and real observations or dynamics. Domain randomization varies simulator properties during training so the model learns behavior that works across a range of conditions instead of depending on one virtual world.
5 min read Explore lesson ↗Technology / NEW ADDITIONPasskeys: how a login proves possession without sharing a password
Learning objective: trace a passkey login, explain why a copied login page cannot normally reuse its proof, and identify the remaining recovery risks. A passkey is a public-key login credential managed by an authenticator, such as a phone, computer or security key. It lets a service verify a cryptographic proof instead of asking you to send a reusable password.
4 min read Explore lesson ↗AI / NEW ADDITIONDifferential privacy: noise, individual influence and privacy budgets
Learning objective: explain how differential privacy limits the influence of one person’s data, calculate a simple privacy budget and recognize the trade-off between privacy and useful results. Differential privacy is a mathematical guarantee about a randomized process: its possible outputs should have similar probabilities whether a particular person’s protected data is included or excluded.
4 min read Explore lesson ↗Technology / NEW ADDITIONWeb crawlers, robots.txt and search indexing
A web crawler is software that visits URLs automatically. Crawling retrieves a resource; indexing makes information searchable. robots.txt communicates crawl rules to cooperative crawlers, while indexing directives influence search results. Neither replaces an actual access control.
4 min read Explore lesson ↗Advanced
Inspect the mechanisms and trade-offs.Inside a Transformer: attention
Attention forms a weighted combination of information from different positions in a sequence. Transformer architectures use attention alongside other components to construct representations.
2 min read Explore lesson ↗AIAgents, tools and prompt injection
A tool-using agent connects model outputs to actions. Prompt injection occurs when untrusted content tries to redirect a model away from the system’s intended instructions.
2 min read Explore lesson ↗TechnologyConcurrency and serializable isolation
Isolation defines how concurrent database transactions can affect what each other sees. Serializable isolation aims for results equivalent to some serial ordering of successful transactions.
2 min read Explore lesson ↗TechnologyTLS, encryption and trust boundaries
Transport Layer Security protects data in transit between endpoints. It combines cryptographic mechanisms for confidentiality, integrity and authentication of the connection.
2 min read Explore lesson ↗FinanceDiscounted cash flow and valuation
Discounted cash flow, or DCF, estimates present value from expected future cash flows and a discount rate that reflects their risk and timing.
2 min read Explore lesson ↗FinanceOptions, payoffs and leverage
An option gives its buyer a contractual right, rather than an obligation, to buy or sell an underlying asset on specified terms. The buyer pays a premium for that right.
3 min read Explore lesson ↗AIDiffusion models and iterative denoising
A diffusion model learns to reverse a noise-adding process. Generation starts from noise and uses learned denoising steps to produce a sample.
2 min read Explore lesson ↗AIPreference alignment: RLHF and DPO
Preference alignment adapts model behavior using judgments about preferred outputs. RLHF uses a reward-based training stage; DPO directly optimizes an objective from preference pairs.
2 min read Explore lesson ↗AIQuantization and deployment trade-offs
Quantization represents model values with fewer bits or a smaller set of numerical levels, often reducing memory use and changing computational requirements.
2 min read Explore lesson ↗AIKnowledge distillation and smaller models
Knowledge distillation trains a student model using signals from a teacher model, such as predicted probabilities or generated examples.
2 min read Explore lesson ↗AICausal inference versus prediction
Causal inference studies how outcomes would change under interventions. Predictive association alone does not identify an intervention’s effect.
2 min read Explore lesson ↗AIDistribution shift, drift and production monitoring
Distribution shift occurs when deployment data differs from the data used to develop a model. Drift describes changes over time in inputs, targets or their relationship.
2 min read Explore lesson ↗TechnologyIdempotency, retries and side effects
An operation is idempotent when repeating it has the same intended effect as performing it once. Idempotency keys help safely retry some state-changing requests.
2 min read Explore lesson ↗TechnologyDistributed consensus and replicated state
Consensus protocols let distributed participants agree on decisions despite specified failures. They underpin some replicated logs and coordination systems.
2 min read Explore lesson ↗TechnologyEventual consistency and user expectations
Eventual consistency means replicas converge after updates stop and propagation completes, under the system’s stated assumptions. It does not promise immediate agreement.
2 min read Explore lesson ↗TechnologyZero trust and least privilege
Zero trust treats access as a decision to verify for a particular request rather than granting broad trust based only on network location.
2 min read Explore lesson ↗TechnologyObservability, metrics, logs and traces
Observability is the ability to investigate system behavior from its outputs. Metrics summarize measurements, logs record events and traces follow work across components.
2 min read Explore lesson ↗TechnologyWebAssembly and portable execution
WebAssembly is a portable binary instruction format designed for efficient execution in a constrained runtime, including web browsers.
2 min read Explore lesson ↗FinanceDuration, convexity and bond sensitivity
Modified duration approximates a bond’s percentage price response to a small yield change. Convexity describes curvature that the first-order duration estimate omits.
2 min read Explore lesson ↗FinanceFutures, hedging and basis risk
A futures contract creates an obligation tied to an underlying reference under standardized terms. A hedge offsets an existing exposure but may leave basis risk.
2 min read Explore lesson ↗FinanceOption Greeks: local sensitivities
Option Greeks describe model-based sensitivities of option value to inputs such as underlying price, volatility and time.
2 min read Explore lesson ↗FinanceCAPM, beta and model assumptions
The Capital Asset Pricing Model links expected excess return to market beta under restrictive assumptions. Beta measures sensitivity to a specified market benchmark.
2 min read Explore lesson ↗FinanceValue at Risk, expected shortfall and stress tests
Value at Risk estimates a loss threshold at a stated confidence level and horizon. Expected shortfall examines average losses in the tail beyond a specified threshold.
2 min read Explore lesson ↗FinanceTokenization, ownership and settlement risk
Tokenization represents claims or records using digital tokens. The legal rights and settlement arrangements behind the token determine what the holder actually owns.
2 min read Explore lesson ↗AIIdeal-point and outranking methods: choosing without hiding trade-offs
Ideal-point methods compare alternatives with a defined best or reference profile. Outranking methods assess whether evidence supports preferring one alternative to another, potentially allowing incomparability.
3 min read Explore lesson ↗AIOptimization under uncertainty, chance constraints and recourse
Optimization selects decisions under an objective and constraints. With uncertainty, a model must also specify distributions or scenarios and when decisions can adapt to new information.
3 min read Explore lesson ↗AIInterpretability, transparency and post-hoc explanations
Interpretability is not one universally agreed property. Transparency about a model’s operation, an explanation of a particular output and a justification for a decision are different goals.
3 min read Explore lesson ↗AIResearch integrity, publication bias and the Chrysalis effect
Research integrity preserves an honest relationship between questions, methods, evidence and reported conclusions. Selective reporting can make a literature appear more decisive than the underlying research.
3 min read Explore lesson ↗AIAI, transaction costs, expertise and institutional power
Transaction costs are the effort and resources involved in finding information, negotiating, coordinating, checking and enforcing exchanges. AI can redistribute these costs and the power associated with them.
3 min read Explore lesson ↗AIIntellectual property, hallucination and managerial responsibility
Responsible AI use distinguishes permission to access material, permission to reuse it and responsibility for the decisions or publications built from it.
3 min read Explore lesson ↗AIProject risk, earned value, performance and closure
Risk management plans for uncertain effects on objectives. Performance control compares evidence with a baseline; closure verifies acceptance, handover and unresolved obligations.
3 min read Explore lesson ↗Finance / NEW ADDITIONProject finance: SPVs, non-recourse debt and DSCR
Learning objective: explain how a large infrastructure project can borrow against its own future cash flows, calculate its debt service coverage ratio, and identify what a lender is actually relying on. Project finance funds a specific asset through a special purpose vehicle, or SPV: a legally separate company created for that project. Limited- or non-recourse debt means lenders rely primarily on the project’s assets, contracts and revenue rather than an unlimited claim on the sponsor.
4 min read Explore lesson ↗