Daily intelligence

Daily Intel: The "Dark Start" Scenario & Starlink Chaos

Published Intelligence Summary & Geospatial Analysis
Report number
558

At a glance

Counts of the stored claim records and follow-up, not a summary of the narrative. How to read these labels

Coverage window
Not stored for this report.
Follow-up investigation
None stored for this report.
Claim records by status
This report has no claim records.
Claim records by confidence
This report has no claim records.
SAR measurements
None supplied to this report. No SAR measurements for this period does not mean that monitored sites were unchanged.
Frontline map data
Age of the map data is not available.

Read with these limits in mind

  • This report is machine-generated from open sources and may contain errors.
  • A source’s publication time is not the time of the event it describes.
  • Frontline map data can lag the reporting period; map dates do not establish when territory changed.
  • No SAR measurements for this period does not mean that monitored sites were unchanged.

Intelligence Summary

Machine-generated narrative. Labels inside it are written by the model. How to read these labels

It’s -27°C, and the war has shifted into a brutal grind against infrastructure. After a barrage of more than 400 drones and 40 missiles, Ukraine’s energy grid is barely hanging on. The Burshtyn Thermal Power Plant in the west is down, severing a vital link to the European grid, and the nuclear plants have tripped offline to save the system. It feels like a deliberate attempt to freeze the population by isolating the western "safe zone" from the national power supply.

To make matters worse, the digital front is a mess. Ukraine tried to update their Starlink "White List" to brick terminals captured by Russians, but the move backfired. Now, frontline units are dealing with connection outages and "blue-on-blue" technical friction right when they need command and control the most.

Despite the gloom, there is some movement on the ground. The International Battalion managed to clear Zoloty Kolodyaz in the Donetsk sector, capturing 18 prisoners—proof that tactical wins are still possible even when the strategic picture looks grim. Meanwhile, in Lyman, the Russians are adapting to the drone threat by deploying "Kurier" ground robots for logistics. It makes sense; why risk infantry running ammo when FPVs are everywhere?

There is a lot of noise coming out of the Dnipro/Zaporizhzhia junction. Russian channels are shouting about a "total collapse" of Ukrainian lines, but I’d take that with a huge grain of salt. It looks like disinformation timed to exploit the Starlink blackout, though they are definitely probing to see if Ukrainian eyes are closed.

Looking at the satellite data, the activity spikes we saw earlier at Site 105 and the 260th GRAU Arsenal have dropped to zero. Usually, when a depot goes quiet like that, it means the munitions have already moved to the launch platforms—which explains the missile wave we just saw. We should keep an eye on the 183rd Guards Regiment, though; their activity is volatile, suggesting they are either reloading or actively maneuvering.

In terms of confidence levels, the grid collapse and diplomatic tensions with Hungary are confirmed facts. However, reports of a breakthrough at Popovka in Sumy come from a single source, so I'm skeptical.

The next 12 hours are a race against physics. If the grid requires a "dark start" in these temperatures, pipes will burst, and the habitability damage will be long-term. If the Starlink outages aren't fixed rapidly, the Russians might try to turn that Dnipro psy-op into a reality. The timing of these strikes suggests they are being used as leverage for whatever negotiations are happening in the background.

Legacy provenance

This archived report predates claim-level evidence records. Treat its confidence labels as editorial assessments rather than independently verifiable citations.

Frontline context

Territorial control and recent changes shown by DeepState.

Report 07 Feb 2026, 15:02 UTC
Map revision comparison

Against the previous report from 06 Feb 2026, 15:03 UTC, bright red is newly marked Russian-occupied and green is no longer marked occupied. For the infiltration area, yellow is this report and cyan is the previous report.

Compare on map
Frontline change measurements are unavailable for this briefing.

Report Visuals

4 visuals
Open full size
Nightwatch daily report #558: Lyman area

Geospatial Analysis

7 locations identified

Historical reconstruction from source editions available by report time.

Places and data export

Coordinates come from geocoding place names found in the report. They locate the named place, not the event described there. Latitude and longitude are WGS 84 decimal degrees.

Show 7 places with coordinates
Geocoded places in report 558
PlaceCountryLatitude, longitudeCopy
Burshtyn Thermal Power PlantUA49.2069743, 24.6666009
Zoloty KolodyazUA48.015883, 37.80285
LymanUA48.9897585, 37.8049964
Dnipro/Zaporizhzhia Operational JunctionUA47.8786956, 35.0700705
Site 105UA0, 0
260th GRAU ArsenalUA0, 0
PopovkaUA51.2188938, 33.117471

How to read the labels

Labels on this page come from three separate mechanisms. A label from one does not mean the same as a similar word from another.

1. In the narrative

Machine-generated narrative. Labels inside it are written by the model.

Unconfirmed:
Written by the model. Its instruction is to mark claims that rest on a single source.
Confidence: HIGH / MEDIUM / LOW
Written by the model. The report stores no definition of these levels, and they are not computed from the claim records. Do not read them as the record confidence above.

2. On claim records: confidence

Record confidence is a threshold on a number stored with the source observation. It is not a probability that the event happened. The follow-up investigation may lower it and never raises it.

HIGH confidence
The status is Independently corroborated, Geolocated or Imagery corroborated, and the stored confidence of the supporting observation is at least 0.75.
MEDIUM confidence
The stored confidence of the supporting observation is at least 0.5 and the rule for HIGH is not met.
LOW confidence
The stored confidence of the supporting observation is below 0.5.

2. On claim records: status

Machine-selected records from the source ledger, labelled by fixed rules. They are a separate selection and are not a list of the sources behind the narrative above.

Single-source claim
One retained source observation supports the record, from a sender that is not on the list of official bodies.
Official claim
One retained source observation, from a sender whose name matches a keyword list of official bodies. It says who made the claim, not that the claim is verified.
Multiple reports, same origin
Several retained observations that do not meet the rule for independent corroboration. They are treated as one origin.
Independently corroborated
At least two distinct source documents from at least two differently named senders. The rule compares sender names and document fingerprints. It does not establish that the senders had separate access to the event, and a repost under another name can satisfy it.
Geolocated
One retained observation carrying coordinates with a geolocation confidence of at least 0.5. It concerns the location only, not the cause or the outcome.
Imagery corroborated
Imagery supports the observation. It does not by itself establish cause or intent.
Analyst assessment
A Nightwatch assessment rather than a directly observed fact.
Contested
Set by the follow-up investigation: material source accounts remain in conflict.
Corrected
Set by the follow-up investigation: the record was changed after the investigation reached a definitive answer.
Insufficient evidence
Set by the follow-up investigation: the available evidence does not support the record as first labelled.

3. In the follow-up investigation

Machine-run follow-up on one selected claim record. Sources listed include material that was retrieved and not used.

Confirmed
The investigation reached a definitive answer supporting the selected claim. The report was rewritten after it.
Refuted
The investigation reached a definitive answer against the selected claim. The report was rewritten after it.
Materially refined
The investigation reached a definitive answer that changes the selected claim in a material way. The report was rewritten after it.
Inconclusive
The investigation did not reach a definitive answer. The report text was kept; the investigation could only attribute, hedge or lower confidence. Individual findings inside it can still carry their own outcome.

Each finding has its own outcome and a confidence written by the model. That confidence is in the finding as worded, which may itself be a statement that something is unknown.

Kinds of investigation source

External web page
A web page retrieved during the follow-up investigation. Its address is stored and linked.
Source observation behind a claim record
A source observation the claim records were built from. No address is stored for it.
Nightwatch archive item
An item found by searching the Nightwatch archive during the investigation. It may be unrelated to the question. No address is stored for it.