SYS NOMINAL FREQ 50.012 Hz LOAD 31.4 GW WIND 8.6 GW SOLAR 2.2 GW NOTES-LIVE 3 UTC 16:08:44
F⌁FATHOM TECHNOLOGYEnergy · Dispatch · Intelligence

Writing from the FATHOM engineering desk.

We write about what we are actually building — methodology, calibration figures, and the failure modes most vendors do not publish. Subscribe if you care about how the work gets done. Roughly one piece a month from the engineering desk. No tracking, no marketing automation, no "executive summary on request." If a piece is online, the whole piece is online.

CADENCE ~1 / month AUDIENCE OPERATORS · ENGINEERS POLICY FULL TEXT · NO GATING FEED RSS & EMAIL
BUS-60

Recent essays

// in reverse chronological order
FN-03 · 2026-04

Sub-second load forecasting is here. Most utilities are not ready for what it changes.

// 2,800 words · forecasting

Forecasting has moved from day-ahead spreadsheets to streaming, probabilistic, substation-level models. We explain what the new operating envelope looks like for control rooms, traders, and procurement teams — what the latency buys you, how the quantile output should be consumed, and what existing HMIs need to change to display it. Includes the ERCOT pilot figures (4–7% additional arbitrage margin on the same battery) and a section on what the next 18 months of DSO-as-forecast-service looks like.

read essay →

FN-02 · 2026-03

Stacking revenue on a grid battery: what actually works in 2026

// 3,100 words · dispatch

Every battery vendor promises stacked revenue. Most operators leave 25–40% on the table because naive stacking — wholesale, then frequency response on residual capacity, then capacity market on top — does not survive contact with reality. We walk through the four real revenue stacks, where they conflict, and the optimisation rules that resolve them. The +14.2% uplift we publish is calibrated against two independent counterfactual methods; both are documented here.

read essay →

FN-01 · 2026-02

Predicting transformer failure 19 days early: the model behind FATHOM Sentinel

// 2,600 words · sentinel

Transformer failures are the most expensive grid events that nobody talks about. A medium-voltage transformer failure costs £180k–£420k all-in, and most of that cost is reactive — replacement under emergency conditions, with crew, transport and outage costs that dwarf the asset itself. We describe the multi-modal model behind FATHOM Sentinel (graph transformer fused with 1-D conv tower on PD waveforms), the 14-month UK DNO trial, and the data discipline required to make it useful in the field.

read essay →

BUS-61

What we write about, and what we do not

// editorial policy

Three rules govern what lands on this page. Rule one: every figure is a calibrated production number or it does not appear. If a number comes from a lab calibration or a benchmark window we say so explicitly. Rule two: every essay names at least one thing we got wrong, or one thing the approach cannot do. If a piece reads like marketing it does not ship. Rule three: nothing is gated. The full text is on this site; the email subscription is for cadence, not for content access.

What we will not publish: customer telemetry without explicit written consent; revenue figures attributable to a single named asset; vendor comparisons where we have not independently re-run the alternative; or speculation about competitor roadmaps. The boundary between writing usefully about the field and being indiscreet about customers is one we take seriously.

SUBSCRIBE

Roughly one piece a month. We send a plain-text email with the title, a paragraph summary and the link. No tracking pixels. Unsubscribe is one click and we honour it the same minute.

RSS

https://fathom-energy.com/field-notes.xml

SIGNAL · NEXT

Want the next essay in your inbox?

If your team would benefit from an early read on the next piece (for example, where it touches a market product or asset class you operate), we sometimes share drafts under NDA in exchange for comments.

open contact form →browse docs →