Meet Athena

Athena explains the hard stuff until it sticks.

Built from primary sources — the actual papers, specs, and people who built these systems. Not generic answers. Finds your gaps, adapts to how you think.

Distributed Systems · Databases · Networking · Consensus · OS Internals

Sift
Sift · Distributed Systems · Raft Consensus
Voice
Lesson 3 of 8 · Raft Consensus
tap mic to speak<400ms
The curriculum

Built from papers and docs. Not generated on the fly.

01
Why Consensus is HardHand-picked

Explain CAP theorem and why two nodes can never fully agree during a partition.

9 min
02
Consensus for the Backend EngineerCustom-built

Map Raft and Paxos onto Kafka, Postgres replication, and Redis Sentinel.

8 min
03
Raft Leader ElectionHand-picked

Step through an election round and explain what prevents two leaders at once.

7 min
04
Clock Skew & Happens-BeforeHand-picked

Use logical clocks to reason about event ordering without trusting wall time.

8 min

Don't see what you need? Ask Athena — custom lessons are scoped to your exact gap, not a generic overview.

After every session

Your knowledge, packaged and ready to go.

When the session ends, Sift doesn't just close the tab. It generates a custom study plan and a Claude context file — so you can keep the conversation going anywhere, any time.

Custom study plan

Generated from what you struggled with

Write-Ahead Logging — durability basics
WAL sequence: append → flush → apply
Checkpointing and log truncation
ARIES recovery algorithm
WAL vs shadow paging
2 of 5 complete

Claude context file

Drop it in any Claude conversation

## Session context — WAL & Durability # generated by Sift · 2026-04-24 Learner understood: - WAL append-before-write pattern - Why sequential I/O is faster than random Gaps to address: - Checkpointing (asked twice, unsure) - ARIES (not yet covered) Suggested next prompt: "Explain ARIES as if I already know WAL"

Paste this into Claude, ChatGPT, or any AI — it already knows where you left off.

Sift · active session
Sift
Sift

So when a leader crashes in Raft, followers notice the missing heartbeat and start a new election. The first one to timeout becomes a candidate.

Sift
Sift

Quick check

What stops two candidates from winning the same election at the same time?

A
The timeout? Each node waits a bit before voting?
Sift
Sift

Close — the randomised timeout helps candidates startat different times. But the real guard is majority voting: a candidate can only win if more than half the cluster votes for it. Two candidates can't both get majority from the same set of nodes.

Adding majority voting to your study plan — it was the missing piece here.

Not just explaining — testing

It checks if you actually got it. Then fills the gap.

After every concept, Sift asks a targeted question — not a quiz, but a real probe. It listens to your answer to find the exact point where your mental model breaks down, not just whether you got it right or wrong.

When it finds a gap, it doesn't repeat itself. It reframes, uses a different analogy, or backs up one level — whatever closes the gap fastest.

Targeted questions

One focused probe per concept, not a generic quiz.

Gap detection

Finds the exact flaw in your reasoning, not just a wrong answer.

Adaptive follow-up

Different angle, different analogy — until it clicks.

Gets smarter with every session

A teacher that learns how to teach you.

Every session teaches Athena something. Which analogies land for backend engineers. Where visual learners get lost. Which questions expose a gap versus which ones just confirm what someone already knows.

More sessions → sharper gap detection → better explanations → more sessions that work. The loop compounds. By the time Athenatalks to you, it's already absorbed what worked for the thousands of engineers who came before.

Why not just ask ChatGPT?

You can — but a general model has no memory of where your mental model breaks down, no lesson graph that sequences concepts correctly, and no data on which explanations actually work for engineers with your background. Athenaisn't smarter. It's been trained on the specific problem of teaching this, to people like you.

Same concept, adapted to who's asking

A
Backend engineer, knows TCP

“Think of WAL like TCP's write buffer — you acknowledge the packet before it hits the application. Same idea: commit to the log, confirm durability, then apply asynchronously.”

B
CS student, new to systems

“Imagine writing in a notebook before updating a whiteboard. If someone erases the whiteboard mid-write, your notebook still has everything. That notebook is the WAL.”

Sift inferred the right framing from your background — no profile setup required.

Lessons designed by engineers from

Google
Meta
Netflix
Stripe
Cloudflare
Figma
Shopify
GitLab
Atlassian
Linear
Google
Meta
Netflix
Stripe
Cloudflare
Figma
Shopify
GitLab
Atlassian
Linear

Trusted by engineers preparing for and working at

Apple
Spotify
Airbnb
Notion
Dropbox
Vercel
Discord
Airtable
Asana
Uber
Apple
Spotify
Airbnb
Notion
Dropbox
Vercel
Discord
Airtable
Asana
Uber
Pricing

Cheaper than your morning coffee.

No commitments. Start free, upgrade when you want more.

Free

$0/ month

No card required. Just start.

  • 1 hour of learning with Athena monthly
  • Hand-picked & custom lessons
  • Visual diagrams
  • Session study plan
Get started free
Coming soon

Premium

$5/ month

Less than a cup of coffee.

  • Unlimited learning with Athena
  • Everything in Free
  • Claude context file export
  • Adaptive study plan that evolves
  • Priority access to new topics
What people say

Engineers who finally got it.

I'd been Googling WAL for two years. Athena explained it in 8 minutes and I actually got it. Got the offer two weeks later.

P

Priya S.

Software Engineer

The gap detection is uncanny. It asked me one question and immediately knew I'd confused linearizability with serializability. No textbook ever caught that.

J

James K.

Backend Engineer

I joined a new team working on distributed infra. Used Athena to get up to speed in a week instead of the usual month of confusion.

A

Anika R.

Infrastructure Engineer

I've watched hours of conference talks on Raft. Athena's questions in 10 minutes did more for my actual understanding than all of them combined.

M

Marcus T.

Senior Engineer

Asked Athena why my Postgres queries slowed after 1M rows. It built a custom lesson on the spot. That's never happened with any course platform.

D

Divya M.

Backend Engineer

The voice mode feels like talking to a staff engineer who has infinite patience. I can think out loud and it picks up exactly where my reasoning breaks.

C

Chen W.

Software Engineer

System design interviews used to feel like memorizing patterns. Now I understand why the patterns exist. Big difference.

O

Oluwaseun A.

Software Engineer

I used the Claude context file to continue a Raft session three days later. It remembered every gap I had. No other tool does that.

S

Sophie L.

Infrastructure Engineer

I prepped for a distributed systems round in 4 days with Athena. The adaptive explanations are real — it noticed I think in terms of code, not theory, and adjusted.

R

Rahul P.

Staff Engineer