See what it does
Each recording is summarized twice: once from the sound alone, the way transcript tools work, and once by Screen Gist, listening and watching. Highlighted details were only ever on screen.
A tech lead and a senior engineer on a live incident call. The engineer shares a dashboard, a terminal on the broken server, the fix and the incident doc. They point at the screen ("that panel", "the line at the bottom") instead of reading numbers out.
Listening only is what a transcript-based summary gets. Listening and watching is Screen Gist. Highlighted: only on screen, never said.
Listening only
0 of 11 on-screen details
A morning spike in 502 errors on the API was traced to two servers whose disks filled up because debug logging had accidentally reached the production config. Alex truncated the logs to restore service, a config fix was shipped, and follow-ups were assigned, with the status page to say it was a resolved configuration issue with no data lost.
- 0:02Incident overview — Speaker 1 needs a status page update in ten minutes, and Alex shares the API dashboard showing errors climbing after nine.
- 0:29Two affected hosts — Only two of six servers show high load, and both filled their disks since Monday.
- 0:58Why disks cause 502s — Nginx can't buffer large responses to disk and returns 502s, so large requests fail while small ones succeed.
- 1:15Debug logging fills the disk — The application log is huge because debug logging writes full payloads, and weekly rotation with no size limit didn't help.
- 2:11Root cause in the release — A staging debug log-level change for a timeout bug rode along into Monday's production config and was missed in review.
- 2:42The fix and recovery — The log level goes back to info, rotation becomes daily with compression and a size cap, and the current logs are truncated, which brings errors down.
- 3:41Timeline and impact — The incident doc covers the timeline from the first alert after nine to full recovery just after half past, with large responses like audit history and exports affected and checkout up.
- 4:11Follow-ups and status — Four follow-ups are listed (config fix, disk alert, CI check, postmortem draft by Thursday), Speaker 1 takes the alert, and the status page will say it was a resolved configuration issue with no data lost.
Screen Gist, listening and watching
9 of 11 on-screen details
An incident bridge for INC-2307 covers 502 errors on api.bramblecart.com, caused by full disks on two of six API hosts after debug logging reached production with Monday's release. Alex Rivera truncated the logs to restore service, shipped a config and log rotation fix, and agreed follow-ups and a status page message with Marcus Hale.
- 0:00Incident overview — Alex shares the dashboard showing a 5xx spike on api.bramblecart.com, and Marcus needs a status page update within ten minutes.
- 0:29Two hosts failing — Only api-prod-2 and api-prod-5 are failing, because their disks filled around 09:12 and Nginx returns 502 for large responses it can't buffer.
- 1:15Root cause in the logs — On api-prod-5, a 94G checkout-api app.log fills /var because debug logging writes full payloads, and logrotate was weekly with no size cap.
- 2:11How debug got enabled — PR #2214 shows a staging debug setting rode into the production config with Monday's release (v2.31.0) and was missed in review.
- 2:42The fix — The PR sets the log level back to info and moves rotation to daily with compression and a 2G size cap.
- 3:02Immediate recovery — Alex truncates the current log on both hosts after copying off the last hour, and errors and disk usage drop.
- 3:41Timeline and impact — The incident doc lists 1,860 failed requests over 19 minutes (09:12 to 09:31), mostly large responses, with checkout staying up and no data lost.
- 4:11Follow-ups and status — Follow-ups are the config fix, a disk alert (Marcus takes it), a CI check for debug logging in production, and a postmortem draft by Thursday; the status page will say it was a configuration issue on two servers, resolved, with no data lost.
Detail by detail
Things shown on screen but never said out loud. Did each summary get them?
| Shown on screen | Listening only | Listening and watching |
|---|---|---|
| Incident number (INC-2307) | No | Yes |
| Which servers failed (api-prod-2, api-prod-5) | No | Yes |
| When the disks filled (09:12) | No | Yes |
| Size of the runaway log (94G) | No | Yes |
| The release that turned on debug logging (v2.31.0) | No | Yes |
| The fix's pull request (#2214) | No | Yes |
| The new log size cap (2G) | No | Yes |
| Failed requests (1,860) | No | Yes |
| How long it lasted (19 minutes) | No | Yes |
| Peak error rate (38%) | No | No |
| Disk alert threshold (80%) | No | No |
A made-up call recorded for this page, with synthetic voices; the company, people and numbers are invented. Both summaries are Screen Gist's own output on the same recording, unedited apart from formatting.
A head of analytics walks a board through a grocery chain's year in 15 slides. The narration keeps to the headlines ("revenue grew nine percent", "almost a third"); the exact figures are only on the slides.
Listening only is what a transcript-based summary gets. Listening and watching is Screen Gist. Highlighted: only on screen, never said.
Listening only
0 of 12 on-screen details
Dana presents Hearthfield's year in three minutes: revenue grew 9% and margins improved, driven by fresh food, online and own brands, while waste and the north region underperformed. She asks for approval of three moves for next year and two decisions today, on the refit budget and the January markdown pilots.
- 0:01Introduction — Dana frames the headline review, noting the full pack went out on Friday and that three moves and two decisions will be requested.
- 0:32Revenue growth — Revenue grew 9%, every quarter beat last year's, and the holiday quarter was the best ever.
- 0:44Sources of growth — Fresh food led growth (produce and bakery fastest), frozen declined, online rose after same-day delivery launched, own brands reached almost a third of sales, and loyalty members spend over twice as much as others.
- 1:40Trip frequency — Customers spend more per trip but visit slightly less often, which the team wants to see reverse.
- 1:51Where money was lost — Fresh food waste ran well above plan, worst in the north, which was the only region to miss plan and has some loss-making stores.
- 2:18Supply chain — The new distribution center opened in the spring has made deliveries more reliable and shelves fuller.
- 2:28Three moves for next year — Grow own brands, use dynamic markdowns on fresh food, and fix the north by refitting or closing loss-making stores.
- 2:44Decisions for today — The refit budget and the markdown pilots starting in January need decisions, followed by a call for questions.
Screen Gist, listening and watching
12 of 12 on-screen details
Dana Brooks presents Hearthfield Market's FY2026 review to the board: revenue grew 9% to $412M and margin and online share improved, but food waste and the North region lost money. She concludes with three FY2027 moves worth $13.2M a year in operating profit and asks the board for two decisions today.
- 0:00Introduction — Dana Brooks opens the 18 November 2026 board meeting and says she will stick to the headlines.
- 0:15Executive summary — FY26 revenue was $412M (up 9%), operating margin 6.4%, online 18% of sales and 1.24M loyalty members, with three proposals for FY2027.
- 0:32Revenue by quarter — Every quarter beat last year's, and Q4 at $117M was the strongest on record.
- 0:43Where growth came from — Fresh produce (+14%) and bakery (+11%) led, while frozen (−3%) was the only category to shrink.
- 1:01Online growth — Online share rose from 11% to 18% in two years, jumping after the same-day delivery launch.
- 1:15Own brands — Own brands reached 31% of sales, up from 24%, with a gross margin 9 points higher than branded goods.
- 1:30Loyalty members — Members drive 61% of sales, 74% are still shopping after 12 months, and they spend 2.3 times a non-member.
- 1:40Basket size and visits — The average basket rose 7% to $41.10, but visits per customer per month fell 6% to 3.4.
- 1:51Food waste — Waste cost $14.2M, and the North was worst at 4.6% against a 2.8% plan.
- 2:06Regional performance — North is the only region below plan (2.1% margin, +3% growth), and four of its stores lose money.
- 2:18Supply chain — After the distribution centre opened in March 2026, on-time deliveries rose to 96% and out-of-stocks fell from 6.1% to 4.2%.
- 2:28Three moves for FY2027 — Grow own brands to 35% of sales (+$6.1M), use dynamic markdowns on fresh food (+$4.3M), and refit or close 4 North stores (+$2.8M).
- 2:44Decisions needed today — The board is asked to approve the $9M North refit budget and the markdown pilot in 12 stores from 6 January, with owners and due dates shown.
Detail by detail
Things shown on screen but never said out loud. Did each summary get them?
| Shown on screen | Listening only | Listening and watching |
|---|---|---|
| Revenue ($412M) | No | Yes |
| Operating margin (6.4%) | No | Yes |
| Best quarter ($117M in Q4) | No | Yes |
| Fresh produce growth (+14%) | No | Yes |
| Own-brand share of sales (31%) | No | Yes |
| Loyalty members (1.24M) | No | Yes |
| Average basket ($41.10) | No | Yes |
| Cost of food waste ($14.2M) | No | Yes |
| North's waste rate (4.6%) | No | Yes |
| North's operating margin (2.1%) | No | Yes |
| What the three moves are worth ($13.2M) | No | Yes |
| The refit budget ($9M) | No | Yes |
A made-up deck and company built for this page, narrated with a synthetic voice; every number is invented. Photos on the slides are from Unsplash. Both summaries are Screen Gist's own output on the same recording, unedited apart from formatting.
A physics lecture on heat engines, written up on a blackboard as the lecturer talks. He explains the ideas; the formulas, worked examples and figures are only in chalk.
Listening only is what a transcript-based summary gets. Listening and watching is Screen Gist. Highlighted: only on screen, never said.
Listening only
0 of 10 on-screen details
This lecture introduces heat engines, explaining that work output equals heat in minus heat out over a full cycle, so no engine can be 100% efficient. It concludes that the Carnot limit, which depends only on the hot and cold temperatures in Kelvin, caps efficiency, and real engines fall short of it.
- 0:01Introduction — The lecture opens on heat engines and the limit none of them can beat.
- 0:14First law and cycles — Heat in minus work out gives the change in internal energy, and over a full cycle the work equals heat in minus heat out.
- 0:42Engine picture and efficiency — Heat flows from the hot side, part becomes work, the rest goes to the cold side, and efficiency is work divided by heat paid for.
- 1:00No perfect efficiency — Because some heat must go to the cold side, no engine can be a hundred percent efficient.
- 1:13Carnot limit — The best possible engine depends only on the two temperatures, which must be in Kelvin, so even a perfect engine here throws away more than half its heat.
- 1:38Raising the hot temperature — A hotter hot side raises the limit, which is why power stations run steam as hot as they safely can.
- 1:59Real engines — Friction, heat leaking away and working quickly keep real engines below the limit.
- 2:09Homework and close — Students are asked to try problem 4.2 and bring their answer on Thursday.
Screen Gist, listening and watching
9 of 10 on-screen details
This lecture covers heat engines, from the first law and efficiency to the Carnot limit, which no engine can beat. It concludes that no engine can be 100% efficient and that real engines fall well short of their Carnot limits.
- 0:00Introduction — The lecturer opens with heat engines and the one limit none of them can beat.
- 0:14First law for engines — Heat in minus work out gives the change in internal energy (ΔU = Q − W), which is 0 over a full cycle, so work equals heat in minus heat out.
- 0:42Engine diagram and efficiency — Heat QH flows from the hot reservoir, part becomes work W and the rest QC goes to the cold reservoir. Efficiency is η = W/QH = 1 − QC/QH, so no engine is 100% efficient.
- 1:13Carnot limit — The best possible engine depends only on the two temperatures, in Kelvin: ηmax = 1 − TC/TH.
- 1:28Worked example — With TH = 500 K and TC = 300 K the limit is 40%, so even a perfect engine discards more than half its heat. A hotter source of 900 K raises it to about 67%, which is why power stations run their steam as hot as they safely can.
- 1:50Real engines — A car engine reaches about 25% (limit about 60%), a power plant about 45% (limit about 65%) and a combined cycle about 60% (limit about 80%). Friction, heat leaks and finite time cause the losses.
- 2:09Homework and close — Students should try problem 4.2 and bring their answer on Thursday, and the lecture ends.
Detail by detail
Things shown on screen but never said out loud. Did each summary get them?
| Shown on screen | Listening only | Listening and watching |
|---|---|---|
| First law (ΔU = Q − W) | No | Yes |
| Carnot limit (η = 1 − Tc/Th) | No | Yes |
| Example temperature (500 K) | No | Yes |
| Example limit (40%) | No | Yes |
| Hotter source (900 K) | No | Yes |
| Its limit (≈ 67%) | No | Yes |
| Car engine (≈ 25%) | No | Yes |
| Power plant (≈ 45%) | No | Yes |
| Combined cycle limit (≈ 80%) | No | Yes |
| Homework heat input (1,200 J) | No | No |
A made-up lecture built for this page: chalk writing drawn by code, chalk sounds synthesized, narrated with a synthetic voice. Both summaries are Screen Gist's own output on the same recording, unedited apart from formatting.