THE NEXT CHAPTER OF SOFTWARE ENGINEERING
Anyone can generate code now.
Almost no one can direct it.
The complete engineering system for building real software with AI.
Not prompting. Not vibe coding. Real engineering.
Two requests. One slot.
Only one confirmed booking.
THE SIGNAL IS ALREADY HERE
Companies are hiring for agentic workflows. Look past the title: they want engineers who can orchestrate tools, manage context, evaluate outputs, and own delivery.
AWS is hiring engineers for AgentCore Memory: systems that help agents retain context, with production delivery, architecture, testing, and reliability at their core.
The role builds agentic coding tools and calls for tool use, orchestration patterns, developer workflows, and model evaluation.
Build the skills to contribute from Lebanon to international teams: context, orchestration, reliable delivery, and human judgment.
JetBrains reports that 90% of professional developers in its study use coding agents at work at least weekly. 68% use them daily.
Original chart + methodology · JetBrains Nearly a third of daily AI users in the survey spend four or more hours a day using AI tools. The opportunity is learning to use that time well.
Original chart + survey results · Stack Overflow THE COST OF “IT LOOKS LIKE IT WORKS”
The failures are documented. The lesson is practical: an agent’s confidence is never your acceptance test.
An incident involving Jason Lemkin’s app exposed the risk of letting development tooling touch live data. Replit recovered the data and rolled out separate development and production databases.
Separate environments. Restrict access. Test recovery.
Original GitHub issue · user report A user reported file deletion after allowing commands during project setup. A project contributor discussed the confirmation controls, and a follow-up change added a home-directory warning.
Review commands. Isolate work. Keep recoverable checkpoints.
You don’t need an agent that sounds certain.
Scope it. Constrain it. Test it. Review it. Then ship it.
Build that system with usTHE TECHTALKS ENGINEERING SYSTEM
Define the business problem, boundaries, and acceptance criteria before writing code.
Give agents relevant context. Delegate bounded work. Keep decisions and permissions explicit.
Reproduce bugs, run meaningful tests, inspect the diff, and challenge confident answers.
Integrate carefully. Review the PR. Prepare the release and explain what remains uncertain.
REAL WORK. THREE DIFFERENT ANGLES.
All three workstreams run throughout the course. You’ll practice new-product delivery, unfamiliar-codebase work, and reusable agent workflows.
A multi-tenant booking SaaS teaching product. Work on scoped features, permissions, availability, and a test proving two requests can’t confirm the same slot.
Illustrative product preview · scope grows through milestones.Explore an existing scheduling codebase. Investigate curated issues, make narrow changes, and submit PRs to the TechTalks educational fork.
Illustrative workflow · an upstream merge is not a course requirement.Create portable agent skills for repo exploration, planning, debugging, reviews, and testing. Evaluate them on real tasks and document their failure modes.
Course workstream · specific contribution setup is confirmed during the course.8 WEEKS. 16 LIVE SESSIONS. ONE ENGINEERING MINDSET.
From understanding the agent loop to shipping real software. See every topic and its reference duration, with a separate live-session plan for work across our three projects.
8 modules / 42 individual lessons
8 weeks / 16 live sessions
Times are reference video lengths, separate from live teaching, discussion, and project work. Missing timestamps are marked “To verify”.
4 lessons · 19:30 reference video
8 lessons · 21:00 timed so far · 4 to verify
6 lessons · 8:30 timed so far · 5 to verify
4 lessons · Timing pending
4 lessons · 9:30 timed so far · 3 to verify
5 lessons · Timing pending
4 lessons · Timing pending
7 lessons · Timing pending
58:30 of supplied reference timestamps. 10 lessons timed; 32 awaiting verification. Titles and remaining timestamps will be reconciled with the reference screenshots.
WEEK 1 / SESSION 01
Lab 01 — BookFlow PRD and Two-Workflow Baseline
Submit PRD, baseline results and an engineering log; identify failures using the same acceptance criteria.
WEEK 1 / SESSION 02
A01 — Compare Two Agent Workflows
Compare two agent workflows against the same bounded task, document errors and verification.
Lab 02 — Agent Workspace and Repository Setup
Clone/run the Cal.diy educational fork; record setup blockers and a preliminary map; initialize techtalks-skills contribution standards.
WEEK 2 / SESSION 03
Lab 03 — BookFlow Architecture and Engineering Plan
Submit an architecture diagram, decision record and thin-slice backlog with explicit non-goals.
WEEK 2 / SESSION 04
Lab 04 — Project Context and Booking Workflow Trace
Review AGENTS.md/CLAUDE.md; trace UI to API to persistence in Cal.diy with file references; draft project-architect or prd-generator inputs/outputs.
A02 — Context Engineering Pack and Architecture Review
Submit project instructions, scoped context and architecture review with verified repository references.
WEEK 3 / SESSION 05
Lab 05 — First techtalks-skills Contribution
Submit a skill PR with triggers, workflow, examples, limitations and baseline comparison.
WEEK 3 / SESSION 06
Lab 06 — First BookFlow Booking Slice
Implement services, staff basics and a booking path; demonstrate tenant boundaries; reproduce the first curated Cal.diy issue.
A03 — Implement and Verify One Complete Feature Slice
Deliver a narrow end-to-end feature with acceptance criteria, tests and an explainable PR.
WEEK 4 / SESSION 07
Lab 07 — Scoped MCP Integration
Demonstrate minimum necessary permissions, safe credential handling, a failure case and a tool-aware skill revision.
WEEK 4 / SESSION 08
Lab 08 — Tool-Assisted Bug Fix and First PR
Submit the first educational-fork bug-fix PR with reproduction, regression test and action evidence; apply the workflow to a BookFlow feature.
A04 — Tool-Integrated Workflow Report
Report tool permissions, actions, failures and verification evidence from a bounded workflow.
WEEK 5 / SESSION 09
Lab 09 — Bounded Delegation and Availability
Implement staff availability using controlled roles and handoffs; evaluate coordination overhead and conflicting findings.
WEEK 5 / SESSION 10
A05 — Delegated Feature Workflow and Human-Reviewed PR
Submit delegation contracts, consolidated findings and a human-reviewed PR with independent checks.
Lab 10 — Human Review and Authorization
Verify role-restricted management flows; review another student's PR; refine reviewer and testing skills with evaluation cases.
WEEK 6 / SESSION 11
Lab 11 — Cal.diy Architecture and Issue Investigation
Deliver architecture map, failing test and root-cause report for a second curated issue; exercise bug-investigator.
WEEK 6 / SESSION 12
Lab 12 — Legacy Patch and Behavior-Preserving Refactor
Submit a second contribution with constraints and regression tests; fix BookFlow timezone/validation edge cases; extend skill regression fixtures.
A06 — Before/After Refactor With Proof of Behavior Preservation
Submit a bounded refactor with before/after behavior evidence and regression tests.
WEEK 7 / SESSION 13
Lab 13 — BookFlow Reliability and Security Gate
Run unit/integration/end-to-end checks, concurrency cases and authorization audit; document actual outputs and unresolved risks.
WEEK 7 / SESSION 14
Lab 14 — Release Candidate and Operational Readiness
Create deployment candidate, logging/rollback runbook, final PR regression evidence and skills release candidate with compatibility notes.
A07 — Production-Readiness Report and Tested CI Workflow
Document correctness, security, deployment readiness, actual CI outputs and remaining limitations.
WEEK 8 / SESSION 15
Lab 15 — BookFlow MVP Completion and Defense
Demonstrate critical user journeys, mobile behavior and data correctness; prepare README, system diagram and demo rehearsal.
WEEK 8 / SESSION 16
Lab 16 — Final Engineering Showcase
Present deployed BookFlow, tests and security boundaries; show reviewed PR history and skill before/after evidence; document next steps.
Initial plan: two live sessions per week. Live time allocations and cohort dates are pending confirmation. Session capacity will be reviewed without removing lessons or extending sessions silently.
A CLEAR FIT BEATS A BIG PROMISE
YOU’LL FIT RIGHT IN IF…
THIS ISN’T YOUR NEXT STEP IF…
Recommended foundation: JavaScript/TypeScript, modern web applications, terminal use, and running a repository locally. No prior AI/ML research experience required.
AN INVESTMENT IN THE WAY YOU BUILD
The value is in the engineering practice: making decisions, directing agents, checking the result, and putting your name behind it.
A scoped BookFlow deliverable with documented decisions, tests, and a clear demonstration path.
Bounded Cal.diy fork changes, pull requests, review feedback, and the reasoning behind your fixes.
Tested agent skills, repository instructions, and an AI Engineering Log you can apply to future projects.
Eight weeks of showing your work.
Start with a conversation about your goals.
BEFORE YOU MAKE THE MOVE
You practice the complete delivery loop across three workstreams, with 16 live sessions, assignments, PRs, reviews, and a final showcase. You leave with engineering artifacts and workflows you can use again. Speak with our team about your experience and goals to see whether the program is a good fit.
Yes. You should understand programming fundamentals, Git, branches, pull requests, basic debugging, and how to run a repository locally. JavaScript/TypeScript and modern web development knowledge are strongly recommended. No AI/ML research background is needed.
Eight weeks, with two live instructor-led sessions each week: 16 sessions in total. Expect demonstrations, practical assignments, pull requests, reviews, and a final showcase. Contact TechTalks for the intake schedule and expected weekly workload.
You will work across BookFlow, a scoped booking SaaS teaching product; bounded contributions to a TechTalks educational Cal.diy fork; and reusable, tested agent skills. The goal is a reviewed set of deliverables, not every possible enterprise feature.
You can show how you scope a problem, direct agents, test the result, and respond to code review. Your product milestone, reviewed PRs, and engineering decisions give employers concrete work to assess. Hiring depends on your demonstrated skills, experience, and the role.
The core workflow is tool-neutral. Specific lab tools, access requirements, and any separate subscription or usage costs will be confirmed for the intake.
Share your details below. Our team will contact you to discuss your experience, goals, and whether the program fits your needs. If it is a match, we will explain the schedule, fees, and next steps before you decide.
YOUR NEXT ENGINEERING UPGRADE
Learn to direct the work, verify the result, and ship with asystem you understand.
Tell us about your experience and goals. Our team will walk you through the program and help you decide whether it is the right next step.
Share your details to discuss your experience, goals, and the program. No payment is taken on this page.
Have a question? Talk to usLeave your details. Our team will contact you for a program-fit conversation.