Running order & scoping notes

Reconcile before anything goes on a screen. Every figure in this desk is constructed — built to the published shape and magnitude of the Telangana system, not measured from it. A room of SLDC engineers will accept a synthetic figure labelled as synthetic and will not forgive one presented as real. In order of risk: (1) peak demand magnitude and time of day — get the actual 2025-26 peaks and their timestamps, this is the single most likely challenge from the floor; (2) settlement time block — the desk assumes 15-minute blocks, 96 a day, and every deviation figure changes if that is wrong; (3) substation and line nomenclature — KTP-220, NBP-220 and MDK-400 are invented, and format is read before content; (4) the daily report format in Draft it Exercise A is a reconstruction, so ask for a published or redacted TGSLDC daily report and rebuild the exercise around the real headings; (5) DSM regulation and rate references are deliberately not cited anywhere on the page — do not add a rate or a regulation number without confirming the version in force.
The constructed day. All five drills run on Thursday 12 March 2026 — clear, dry, late rabi, ambient 34 °C by 11:00, forecast peak 16,840 MW at block 45 and an evening secondary maximum at block 78. One day, one context, so the room never has to reset.
Open — the seventh desk. The merged deck puts "Superintending Engineers · system planning" and a Network Planning Assistant on Track 2's card. All six desks on the page are SLDC operational positions, so an SE who attends on the strength of that card finds no desk. Either add a seventh desk for network and transmission planning — which needs a fourth Draft it exercise, most naturally a scheme justification note — or take planning off the slide. Not yet decided. Do not leave a Superintending Engineer to discover the gap by attending.
On the control room line. The exact internal classification of the eight Guard it items is to be confirmed with the IT wing at scoping — this applies to Track 3's version of the module as well, where the items are imagery and asset data rather than control-room data. The participant pages say only that the module teaches the instinct rather than the policy; the confirmation is ours to obtain, not theirs to read about.
On the load shape. The late-morning maximum is the single most likely thing to be challenged from the floor, and it is correct — both 2026 records were set around 11:00, driven by rabi agricultural load. If someone insists the peak is in the evening, open screen A on the monsoon day: there it is, and the reason is that rain takes the irrigation load away.
On calibration. One claim in every Question it set holds up. Say so before they start. An engineer who rejects a sound forecast has cost the system as much as one who accepts a bad one, and blanket distrust is the failure mode you are most likely to create by accident today.
On the deviation exercise. The sharpest artefact on the site. The solar series is sitting in the same table the model was given and it tracks forecast within about one percent in every block — so the invented cause is not merely unsupported, it is contradicted by data the model was holding. Make them read the last three columns before they flag anything. Then run the guardrail and let them see the second output: shorter, duller, defensible, and it surfaces the solar figure instead of hiding behind it. The line to say is that a cause stated in a settlement document is an act by a named officer, and it is the sentence a regulator reads first.
On Triage it. All three collapses are correct and the room will score three out of three. That is the setup, not a soft drill — let them feel good about it, then send them to the bin. The suppressed Buchholz stage-1 alarm is unrelated to the disturbance, which is precisely why it matters, and it was sitting in the raw log the whole time. If someone says the answer is to switch suppression off, agree that it would work and then point out that an alarm system which suppresses nothing is an alarm system nobody reads.

Reconcile before anything goes on a screen. Everything on this desk is constructed — built to the shape of a State transmission network, not measured from one. Field engineers read nomenclature before they read content, and a wrong tower ID format discredits a correct finding. In order of risk: (1) tower, line and bay nomenclature — AB-nnn and the line names are invented, so get the real convention for 400 kV and 220 kV routes, tower numbering along a route, and ICT designations; (2) DGA gas thresholds and the interpretation basis — the Score it values are illustrative, so confirm which standard the O&M wing works to and its current threshold table, and do not put a threshold on a screen without that confirmation, because a room of substation engineers knows these numbers by heart; (3) defect vocabulary — ask for one real, redacted inspection report and use its terms; (4) patrol and clearance cycles — the vegetation drill depends on the prescribed line patrol frequency and clearance cycle; (5) statutory clearance distances are deliberately not stated anywhere on the page, and must not be added without confirming the applicable rule.
On the false negative. AB-125 in Check it is the whole reason the drill exists. It returns severity none at confidence 0.31 and is the only row with no colour on it, so a room reading the red items alone will scroll past it. The actual finding now sits in the low-confidence list, collapsed by default — a real interface hides it because it is mostly noise, and it is mostly noise. Do not point at it and do not open the list for them. Let them reveal, let the score say “you did not open the low-confidence list”, and only then say the line: a clean drone report is not an inspected line.
On the false positive. AB-116 carries the highest confidence in the set at 0.94 and is a shadow. If someone objects that high confidence should mean something, agree with them — it does. It means the model is sure about the pattern it matched, not that the pattern is a defect.
On AB-129, the anti-climb device. The finding is real, the confidence is earned, and the grade is wrong. The model looked at the object and graded the object. It cannot see that the tower stands beside a village road and that children climb. This is the one that produces the R grade on the triage take-home: severity is a judgement about context, and context is what a frame does not contain.
On the work order. The invented hot-line permit is the sharpest moment on the page and the one worth slowing down for. It is a safety-critical sentence produced by a drafting tool that was told about one corroded bolt.
On Score it. Run part one live and add the inputs one at a time — the point only lands if they watch acetylene take it from 71 to 43 in a single step. Then part two: score 82, monitor band, nobody looked at it, and it failed. Make them ask why before you open the input panel. The answer is an eleven-month-old DGA sample sitting behind four live inputs, and the rule that comes out of it is the take-home for the whole day: a health score is only as fresh as its oldest input, and the score does not tell you which one that is.
On vegetation. Not a data column — the deck's Module 3. Six segments, and only two of them are sequenced by risk. Segment D is highest risk and foot-access with a stream crossing, so the terrain sets the date and not the score. Segment E is low risk beside a road, and somebody will want to do it because it is easy; that instinct is what the row is testing.
Open — the sixth desk. The merged deck puts a Project Monitoring & Delay-Prediction Co-pilot on Track 3's card for EE/AEE. Project and construction monitoring appears in no drill on any of the four tracks. Either add a sixth desk for it with a Draft it exercise of its own — most naturally a monthly progress note or a delay justification — or take it off the slide. Not yet decided, and handled the same way as Track 2's seventh-desk question. An Executive Engineer who attends on the strength of that card gets a day about substations.
Before delivery. Confirm the internal classification of the Guard it items with the IT wing, as for Track 2. Confirm also whether the organisation already runs drone inspection with an AI finding step, or is considering it — the drill lands differently if they have a live contract, and that is worth knowing before the room tells you.

Read this before you use the running order below. /track-4 is now a document, not a desk. It is a briefing note — a week on a Director's desk, five days of reading, a Which tool panel, three failure modes, and the draft classification table. Its only interaction is “Mark up the file” on each day, and it is built to be printed and left behind as much as projected.
The interactive governance desk the running order below describes — See it, Rank it, Guard it, Test it, Sign it, the board note and the four take-homes — is not at that URL any more. It is preserved in the repository at commit 1245598 and can be restored or given its own route. Do not walk into the room on the strength of the rail below until the day's format is settled and this section is re-drafted.

Live or canned — decide at scoping, set it on the day. See it is built to run live and ships the pre-generated brief as the fallback, not the plan. Confirm with the IT wing whether an approved tool can be open on a machine in that room; the difference between watching it happen and being shown that it happened is most of the drill's value. The switch is in the drill itself, marked Facilitator setting, and it defaults to live — set it to pre-generated before the room arrives if there is no tool, not in front of them.
On See it, and stopping. Run the good brief, then stop. There is deliberately no next button and no scrolling until you press through — the room is meant to sit with twenty seconds against forty minutes before anything else happens. Then the second half, then the invented 11 percent, then three sentences and nothing more: it is not a malfunction, every rule you write today exists because of it, nothing after this is abstract. Do not add a fourth sentence.
On the taxonomy. Six failures, each opening to the actual artefact from the track that meets it — the pensioner helpline, the solar cause, T-231, the 95 percent you are about to see this afternoon, leakage, and drift. Open five of them quickly. Open Drift slowly and leave it open: it is the only one with no owner anywhere in the programme, it arrives in year two after the warranty conversation is over, and it ends on four questions that belong to this room and nobody else.
On the horizons. Every Rank it card now carries one on its face. When three are chosen the page states the combined picture, and if all three are long-horizon it says so plainly — nothing visible before month 12, a full budget year with nothing to put in front of the Board. If transformer health scoring is among them, the page raises the question to ask out loud: how is DGA sampled today, and where does the result go? If the answer is a paper register the horizon is not 24 months, it is 24 months after a sensing programme nobody has budgeted.
Two horizons to confirm. Ten of the twelve are taken from the source table. Departmental test question banking and regulatory filing assembly have no counterpart there, so their horizons are derived by analogy — published material, no integration, so 1–3 months. Confirm both at scoping before the room sees them.
On the 95% claim. This is the centrepiece and it is worth the whole afternoon. The proposal states the base rate in section 5 — 380 transformers, 11 failures a year — and the accuracy claim in section 2, deliberately far apart. A model that always answers “no failure” scores 97.1%, which beats the 95% offered. Do not point at section 5. Let them reveal, let the arithmetic block do the work, and only then say the line: accuracy on a rare event measures the rarity. The room will have senior engineers in it who get there unaided, and it is much better coming from one of them.
On the fourth tap. Refusing the fourth use case is the only place in the programme where the tool tells a CMD no. Expect somebody to ask for a fourth slot. The answer is that one IT wing and one budget year is the actual constraint, and a Phase 1 of six is a Phase 1 of none.
On the owner field. It rejects a bare wing name — type “IT wing” and it says so. This is where the day either turns or does not. If the room supplies three wings and looks at you, wait. Do not accept “the IT wing will nominate someone”; that is the sentence this constraint exists to prevent.
Re-verify the legal instruments on the morning of delivery. Not at build time, not in the week before — on the day. This is a standing instruction and it applies to every instrument named on the page. A commencement notification issued overnight makes a printed classification table wrong in a room full of people who will act on it, and the CMD is the one who signs the note that carries it. Five minutes with the gazette before the room arrives.
Before delivery — legal. Every reference on the page was checked and the status is printed on the page itself, including what could not be confirmed. Re-check all of it in the week before the session and again on the morning, because two of them move: the CEA (Cyber Security in Power Sector) Regulations were still draft and not notified at the time of writing (drafts in August 2024 and 6 October 2025), and the DPDP Rules, 2025 run on an eighteen-month phased commencement, so which obligation bites on which date changes with the calendar. If either has moved, correct the page rather than correcting it verbally in the room. The CEA Guidelines, 2021 and the CERT-In directions of 28 April 2022 are settled and unlikely to need touching.
No US frameworks. The page names NIST, FedRAMP and SOC 2 exactly once, to say that none of them binds TGTRANSCO and that a vendor leaning on them has told you something. Keep it that way. Anchoring an Indian utility's data policy in American frameworks is the fastest way to lose this particular room.
On the output. Decision 02 produces the draft classification table that Tracks 1, 2 and 3 have been deferring to. Get it out of the room as a printed sheet before the tea break — that is the deliverable we have been promising in the other three decks, and it is the one thing that must not be lost to a closed tab.
Scoping. Confirm we have sight of the Top 20 Critical Issues and Emerging Technologies map before the day, and that the twelve candidates on the page match how the organisation names them; the room recognising its own document is what makes decision 01 credible. Confirm also whether the IT wing attends with the Directors or separately — if separately, decision 02 has no one in the room who can commit, and the day should be re-sequenced.