Stats

Ratings
Brews
Show

Nordic Light averages 7.4, ultralight 7.2, nordic 7.1, light 7.0, hyperlight 6.2:

Custom range

Filters

Caffeine, Default (caffeinated): ratings leave decaf out. Counts, cost and Caffeinated vs. decaf include it.

All time · Mountain Time

239 brews239 brews .

Matching: 221 bags, 1168 brews

Sweet spots

Sweet spots
  • For each setting, the band your ratings favor for coffees like this.
  • First choice: the same bag brewed both ways.
  • +0.4: beat the same bag's other brews by 0.4.
  • Too few of those: bag averages across similar coffees, widening the group until there's enough.
  • Likely / Possible: the gap is 2×+ / 1.5–2× its margin of error (about 95% / 87% sure).
  • Early: ahead or looking tied, but too few bags to call.
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
90°C or 91°C 93°C looks higher · 2 bags Usually: 90–92°C
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
2 in the running Lotus - Mg/K 60/10 (880mL) looks higher · 1 bag Usually: Lotus - Mg/K 60/10 (900mL)
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
5 in the running Usually: CT62
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
Cafec Abaca or Cafec T90 Usually: Cafec Abaca
Likely
  • The gap is at least twice its margin of error: about 95% sure it isn't chance.
1:15.5–16.5 Usually: 1:16.5–17.5
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
< 16g, 18–20g or 20+g Usually: 18–20g
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
4 in the running Usually: 1.3.0
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
60s+, 30–44s or 45–59s Usually: 60s+
Early
  • Ahead (or looking tied) so far, but too little data to call it.
  • More brews would settle it.
3 in the running Usually: 61–90 d

Top bags like this

Top bags like this
  • The bags matching these filters with the best ratings.
  • Score: each bag's average pulled toward the matching brews' average by 5 brews, so one great brew doesn't top.
  • SD: how much its ratings vary.
  • Small: consistent.
  • Big: hit or miss.
  • Under each: its roaster, then usual temp · water · dripper.
BagBrewsScore
SD · standard deviation
  • How much ratings vary from brew to brew.
  • Small: consistent.
  • Big: hit or miss.
  1. 1 Gachiku Estate TOH #1 Natural SL28 Big Sur · 92°C · Lotus - Mg/K 60/10 (900mL) · Alpha Switch 6 0.27 8.6
  2. 2 Colombia Gesha Washed El Hatillo Luminous · 94°C · Lotus - Mg/K 60/10 (900mL) · Alpha Switch 9 0.65 8.3
  3. 3 Colombia Yoiner Mosquera Gesha Anaerobic Natural Shoebox · 85°C · Lotus - Ultralight Mg + K (900mL) · Graycano 6 0.71 8.2
  4. 4 Masincho Washed 74158 Big Sur · 92°C · Lotus - Mg/K 60/10 (900mL) · Alpha Switch 5 0.35 8.1
  5. 5 La Terrazan Java Bean's Beans · 91°C · Lotus - Ultralight Mg + K (900mL) · CT62 4 1.03 8.1
Built with help from Claude Code, an AI coding assistant How AI was used Contact

How this site was built

I built Bean Tracker with Claude Code, Anthropic's AI coding assistant. Here's who does what.

What I do

  • Set the direction: which features get built, and in what order
  • Drive the technical decisions, drawing on my experience as a Staff SRE, DevOps engineer, and software engineer: architecture, hosting, testing, and releases
  • Bring nearly all of the coffee knowledge, from recipes, gear, and pour structure to extraction, rest, and water
  • Shape every design: I review each spec and push back on layout, wording, and behavior until it's right
  • Test every change in a live preview, on desktop and on my phone
  • Set the engineering standards: tests at every screen size, shareable links for every view, signed commits, and staged releases
  • Own the data: every coffee, rating, and note, plus the judgment calls on it
  • Run the infrastructure the site lives on: hosting, domains, and backups
  • Act as project manager and technical lead, and decide what gets published, including this note

What Claude Code does

  • Writes nearly all of the code and automated tests, following my direction
  • Drafts the designs, specs, plans, and docs for me to review
  • Researches options and explains the tradeoffs so I can decide
  • Analyzes my brewing data to suggest values, which I confirm or change
  • Releases and deploys changes once I give the go-ahead
  • Often runs several AI agents at once: one builds while another reviews

Nothing ships without my sign-off. I review every change in a live preview, and automated tests plus a separate AI code review check the code itself.

Questions or concerns?

If you have a question, spot a mistake, or want something removed (for instance, a photo of your coffee's label), email me at [email protected].

v1.110.7 Last updated: Last deployed: