כל מה שכדאי לקרוא לפני שמתחילים אחד הקורסים שלי, ולחזור אליו תוך כדי העבודה: איך לגשת ללימודים, מאילו הנחות כדאי להיפרד, מה פרויקט טוב באמת צריך להראות, ואיפה למצוא דוגמאות של עבודות קודמות, עוזרי קורס וספרי הלימוד שעליהם הקורסים נשענים. למאמרים ארוכים יותר על AI בהנדסת תוכנה ובאקדמיה, מוזמנים לקפוץ גם לפוסטי הבלוג שלי.
קראו את אלה מוקדם. הם מעצבים איך תעבדו בתוצאות התואר ואיך יישפט כל מה שאתם מייצרים.
Advice to CS Students in 2026
Six pieces of advice for navigating a degree in the AI era, each with a pattern to adopt, an anti-pattern to avoid, and the inner statement that gives the anti-pattern away. Available in English and Hebrew.
The Honest, Brutal Truth About Learning
Decide and shape your skillset, then face five hard truths: aim for skills over passive knowledge, practice relentlessly, do not trust the illusion of mastery school leaves you with, and use AI as your honest mirror.
Will AI Make Software Engineers Obsolete?
Common themes as of July 2026: the changing role of the engineer, where vibe coding reaches its limits, writing much less code, why code is still taught, and how uncertain the future really is.
Work Backward From the Job
Employers hire for specific roles, not for “CS graduates.” Five steps to prepare for the job market before you graduate: pick target roles, learn what employers actually expect, build a learning plan, build a convincing portfolio, and practice defending your work.
התמצאות לסטודנטים בקורסי LLM, GenAI וקורסי AI קרובים: מה מצופה בקורסים האלה, אילו הנחות כדאי להוריד, ואיך להפיק את המרב מהחיבור שבין התואר שלכם ל-AI.
The One Kind of Innovation These Courses Require
The three questions behind any project (research, engineering, product) and the three kinds of innovation that answer them. Explains why these courses target scientific innovation, with product value as a floor.
Ten Things Students Get Wrong About Course Projects
Ten beliefs that quietly limit what students get out of a project-based course, from mistaking a working demo for engineering skill to treating a course as content to consume. Each comes with the sentence that gives it away.
Medical Technology Students: Don’t Waste Your Cross-Domain Advantage
How to turn the intersection of your degree, medicine and AI, into a standout LLM project: choose a meaningful problem, develop deep domain understanding, keep technological depth, and connect both worlds.
The Recordings Are Not the Course
I record the sessions, but watching lectures is the least important part. The value is live interaction and three required project presentations, so the course is not a fit if you only plan to watch the recordings.
Work Smart, Not Long
You are graded on the quality of the result, not on hours invested. Five habits that make the same time produce a better project: work smart, start early, reuse aggressively, use the course material, and validate continuously.
How These Courses Teach Theory Through Tools
Tools (libraries, frameworks, APIs) are integrated into the study of theoretical concepts. Three ways to get more value: look beyond the code to the concept, let tools shape your project design, and build tool experience that matters for your career.
English in Your Studies
The field runs on English. You do not need perfect English, but you do need a basic working level, the standard technical terms, and the habit of working in English from the start.
How to Write to Your Instructor
Five habits for a good message: try to answer it yourself first, include all the context, be brief and careful with wording, stay polite and professional, and expect links or an AI transcript in reply.
Preparing for Your In-Class Presentation
A six-step checklist for the days before you present: set up and register your repository, review the guides, self-evaluate with the course bot, rehearse and time your talk, and act on the instructor's feedback afterwards.
Don’t Let a Bad Talk Sink a Good Project
Ten tips for presenting technical work: build a narrative, justify only the choices that had real alternatives, keep slides clean, speak in precise technical language, rehearse on camera, and spend your time where the information is.
AI Does Far More Than Write Your Code
Five ways to get real value from AI in your projects: invent new uses beyond generating code and slides, own every result, treat AI as a utility, and meet the higher bar it now sets. Each idea comes with a good and a bad example.
Suggested Weekly Plan for Project Execution
A week-by-week roadmap for a 13-week semester: which requirements to advance each week, with the proposal, midterm, and final at weeks 5, 8, and 13, plus weeks 14 and 15 to polish and submit the repository.
מה פרויקט GenAI או LLM צריך להראות, ואיך בונים את הנתונים שמאחוריו. השתמשו בהם בזמן שהעבודה עדיין מתגבשת.
LLM/GenAI Project Requirements
Twelve requirements a strong course project is expected to satisfy, grouped into four phases from framing the problem through to engineering rigor, each paired with the anti-patterns that most often weaken it.
LLM/GenAI Project Assessment Criteria
The five dimensions a project is judged on: domain understanding, depth of solution-space exploration, novelty and reuse, methodological rigor, and how clearly the work is presented and defended.
Synthetic Data Generation for LLM/GenAI Courses
A do-and-don't guide to generating synthetic data for LLM and GenAI projects: research real data first, control what you produce, prototype and debug small, inspect by hand, and compare strategies before scaling.
Practical EDA for Synthetic Datasets
An eight-objective checklist to run before training: sanity, coverage, realism, variability, label quality, leakage, bias and balance, and difficulty, each with the practical EDA that verifies it.
מה פרויקט חוסן בראייה ממוחשבת צריך להראות ואיך הוא נשפט. השתמשו בהם בזמן שהעבודה עדיין מתגבשת.
Project Requirements for Vision Course
What the vision robustness project must do, stage by stage: dataset and task selection, clean baseline, distortion and degradation, restoration and fine-tuning, and documentation, each with its anti-patterns and a suggested weekly plan.
Project Assessment Criteria for Vision Course
The five dimensions a vision robustness project is judged on: study design, measurement rigor, depth of exploration, recovery analysis, and visualization, each with the strong and weak signals graders look for.
מה כל אבן דרך משמעותית צריכה להראות: לא רק שנעשתה עבודה, אלא שהצוות מבין את הבעיה, קיבל החלטות טכניות מנומקות, העריך את הפתרון נכון ויכול להעביר את התוצאות בבירור.
Project Proposal Presentation
Six parts a strong proposal must cover: motivating use case, problem statement, novelty, models and methods, data specification, and an evaluation plan, each with its requirements and the anti-patterns to avoid.
Midterm Presentation
Six parts that show the project has moved from planning to validated work: refinement, previous work, dataset and exploratory analysis, a baseline, initial results with error analysis, and a completion plan.
Final Presentation
Seven parts that make the final technical argument: the refined definition, achievements and contributions, methodology, experimental results, interpretation, conclusions, and demonstrated effort.
GitHub Submission
Fifteen parts for a professional repository: naming, structure, a complete README, documented data and models, reproducible training and evaluation, code quality, and a final submission checklist.
מתודולוגיית ההוראה, דוגמאות עבודה, עוזרי קורס וספרי הלימוד שעליהם הקורסים בנויים.
Free GPU Resources for Students
For most course projects you do not need to pay for a GPU. Your own machine, the HIT lab computers, and free cloud platforms (Colab, Kaggle, Lightning AI, Modal) cover most of what a project needs.
LLM Access for Students
Each project gets a small OpenRouter budget, but you can also use LLMs for free: OpenRouter’s free models or Ollama locally. Plus practical rules to keep your API costs low.
Past Projects
Completed student course projects with their repositories and write-ups. The most reliable way to calibrate scope and quality before committing to your own proposal.
Course Bots
Course-specific AI assistants grounded in the syllabus and materials. Use them to close knowledge gaps the week they appear, and to challenge your own reasoning rather than to generate work.
See It Before You Code It
A curated list of 3Blue1Brown's visual explainer videos on neural networks, transformers, and how LLMs actually work. Each video as its own card, watched in order they build a strong mental model before the code.
ספרי לימוד מקוונים
The open textbooks behind the course series: Language AI, Vision AI, Scalable AI, Temporal AI, and Embodied AI. Readable online, with companion Android builds.
