Mission Makers, anyone planning on leveraging AI tools for missions?

[Jade_Monkey]

I am very excited to build missions for Korea, especially with all the new quality of life upgrades (better airfield taxiing logic etc).

Sounds like the editor will come with new features:

• support for free scripting, meaning we can really customize the events to very narrow scenarios instead of having to create a frankenstein monster for MCUs.
• text to speech of some sort

With the availability of new tools like Claude and ChatGPT since I last built careers, there seem to be a million more options for mission builders to srteamline and enhance the process.

Here are some high level ideas Ive been experimenting with. Once we have a dedicated section for Korea Mission builders I'm happy to share my workflows in more detail:

• Use AI for historical mission research - Get detailed acconts of missions
• Use AI to generate realistic briefings
• AI generated mission maps or images
• Use AI to generate voiceovers for briefings
• Use AI to generate missions in the editor by providing a briefing

Has anyone considered other uses?

[GenMarkof007]

S! @Jade_Monkey

Very interesting thread. I’m also looking forward to seeing what comes next for mission creation.

For now, I’ve been working with ChatGPT to generate some realistic (fictional) mission briefings and background research for my PvP Tank Crew events.

For the creative side, since I’m a graphic designer, I prefer to create most of the visuals myself (promotional material, in-game visuals, etc.).

I’ll definitely follow this thread once you create a dedicated section on the IL-2 forum.

Thanks for starting the discussion.
Cheers, GenMarkof007  

[AEthelraedUnraed]

1 hour ago, Jade_Monkey said:

• Use AI for historical mission research - Get detailed acconts of missions

Unless you've got a transliteration of original documents like RAF ORBs that you could attach to an LLM to search in, this is basically guaranteed to lead to untrustworthy and outright abysmal results. Even if you add original documents, I wouldn't trust AI and always manually validate whatever it says. However, in that case it can at least help you with filtering for interesting missions as well as summarising them.

1 hour ago, Jade_Monkey said:

• Use AI to generate missions in the editor by providing a briefing

Unless someone specifically retrains a specialised AI, this is extremely unlikely to work. Even with an specialised AI, it's likely to only work with short, simple missions.

On the other hand, the following:

1 hour ago, Jade_Monkey said:

• Use AI to generate realistic briefings
• AI generated mission maps or images
• Use AI to generate voiceovers for briefings

are all valid uses.

Edited March 13 by AEthelraedUnraed

[Jade_Monkey]

@AEthelraedUnraedI fully agree with your assessment overall. However, in practical terms there are some advantages that can accelerate the process of mission creation.

 

17 hours ago, AEthelraedUnraed said:

Unless you've got a transliteration of original documents like RAF ORBs that you could attach to an LLM to search in, this is basically guaranteed to lead to untrustworthy and outright abysmal results. Even if you add original documents, I wouldn't trust AI and always manually validate whatever it says. However, in that case it can at least help you with filtering for interesting missions as well as summarising them.

I made a 13 mission Normandy Typhoon Campaign (unreleased for other reasons), that was 100% based on ORBs and historical accounts. It took me an insane amount of time to gather, verify etc, and to be honest ORBs don't contain enough details to cover all aspects of a full mission. You end up making up a large portion of the mission details regardless. In practical terms, AI filling in the unknown with its own guesses is not as bad if we take off the "purist" hat and put on a "practical guy" hat.

 

17 hours ago, AEthelraedUnraed said:

Use AI to generate missions in the editor by providing a briefing

 

I'm not saying that it can build a whole mission right now, but it can be helpful, and who knows . Yesterday I was messing around with claude code reading mission files to learn how to calculate the Xpos and Zpos (coordinates) on the map based on the in- game overlay grid (e.g. 3510-4). You can build a list of cities and points of interest and then use claude to generate missions with the relevant objects for a given briefing. You dont NEED AI for this but it sure speeds up the process. Im sure Patrick Wilson and other folks would have loved to have current tech when they built their tools back then.

Regarding briefings, here are an example of each: briefing doc, Transcript for the audio, and the actual audio (sorry audio is for a different mission than the briefing but you get the idea).

 

 

Mission 08 - Transcript.pdf Mission 08 - Briefing.pdf Mission 10 Audio .7z

[AEthelraedUnraed]

56 minutes ago, Jade_Monkey said:

I made a 13 mission Normandy Typhoon Campaign (unreleased for other reasons), that was 100% based on ORBs and historical accounts. It took me an insane amount of time to gather, verify etc, and to be honest ORBs don't contain enough details to cover all aspects of a full mission. You end up making up a large portion of the mission details regardless. In practical terms, AI filling in the unknown with its own guesses is not as bad if we take off the "purist" hat and put on a "practical guy" hat.

The issue is not the AI filling in unknown gaps; the issue is the AI completely making up stuff while presenting it as historical. AI won't just make up small details that are not in the ORB (e.g. the route flown), but will get important details like the date, target, time, opposition etc. completely wrong.

You can certainly get an AI to come up with missions that have the right aircraft, mission types, opposition etc. for a particular phase of the war. But it cannot reliably provide historical missions that were actually flown. There's nothing inherently wrong with creating missions that are "historically plausible" rather than fully historical, and in fact many mission makers do so, but if you want to create a historical campaign which is what your question specified, AI is not suitable for the job.

The only exception is, as said, if you have machine-readable documentation available (e.g. ORBs) and use the AI to summarise it. As you say, this is an incredibly time-consuming job to do manually. Usually LLMs can do this pretty well, although you always need to be careful. Again, there's nothing inherently wrong with the AI making up unknown details, as long as you are aware what details were made up, and are aware what that means for the historicity of the mission. Especially if you're going to present the mission to the public while advertising it as historical, you have an ethical obligation to do so.

56 minutes ago, Jade_Monkey said:

I'm not saying that it can build a whole mission right now, but it can be helpful, and who knows . Yesterday I was messing around with claude code reading mission files to learn how to calculate the Xpos and Zpos (coordinates) on the map based on the in- game overlay grid (e.g. 3510-4). You can build a list of cities and points of interest and then use claude to generate missions with the relevant objects for a given briefing. You dont NEED AI for this but it sure speeds up the process. Im sure Patrick Wilson and other folks would have loved to have current tech when they built their tools back then.

AI can usually analyse structured files just fine, which is what I understand you did. Generating such files is a whole different story though, especially if there are obscure links and complex logic across thousands of objects, such as in the case of IL2 missions. At this point in time, AI is simply not up to the task. And I don't see that change in the near future.

Edited March 14 by AEthelraedUnraed

[Juri_JS]

I really don't recommend using AI for historical research. I've tested it with Grok when I did the research for my planned Yak-9P campaign and more than once I caught Grok just inventing events. For example the AI invented a Yak-9P attack on an American destroyer. Because I had read Dilby's works on the North Korean Air Force, I knew it couldn't have happened, but at first it appeared very convincing. When I asked more detailed questions about the event, Grok even produced fake quotes and linked real documents as sources. When I checked these documents I couldn't find the quotes. In the end Grok admitted, that it was all made-up and of course he was really sorry.

[Jade_Monkey]

@AEthelraedUnraed I'm not arguing for a one shot perfect mission generation in chatgpt. I'm just saying there are way to harness AI to help streamline some of the hardest parts of mission building.

 

I'm gonna leave it at that. No wonder I didn't visit the forum in two years. It's exhausting arguing with one-uppers online.

Edited March 14 by Jade_Monkey

[Jade_Monkey]

9 minutes ago, Juri_JS said:

I really don't recommend using AI for historical research. I've tested it with Grok when I did the research for my planned Yak-9P campaign and more than once I caught Grok just inventing events. For example the AI invented a Yak-9P attack on an American destroyer. Because I had read Dilby's works on the North Korean Air Force, I knew it couldn't have happened, but at first it appeared very convincing. When I asked more detailed questions about the event, Grok even produced fake quotes and linked real documents as sources. When I checked these documents I couldn't find the quotes. In the end Grok admitted, that it was all made-up and of course he was really sorry.

I wouldn't recommend asking AI for research either, I'm aware AI can hallucinate. Gather the research materials and ask it to organize them and structure to your liking. It's a lot easier/faster to review that than doing all the work yourself.

Using NotebookLM or Claude projects is pretty effective at it, you can ask for citations to your source files (file, chapter, page# ) to expedite the process.

[Juri_JS]

What worked quite well was using AI to create the artwork for the campaign. I've uploaded the Yak-9Ps images the devs have posted and used them to create North Korean style propaganda posters.

1 minute ago, Jade_Monkey said:

I wouldn't recommend asking AI for research either, I'm aware AI can hallucinate. Gather the research materials and ask it to organize them and structure to your liking. It's a lot easier/faster to review that than doing all the work yourself.

Using NotebookLM or Claude projects is pretty effective at it, you can ask for citations to your source files (file, chapter, page# ) to expedite the process.

I might give it a try. Could you give a more detailed explanation how to do it.

[Jade_Monkey]

51 minutes ago, Juri_JS said:

I might give it a try. Could you give a more detailed explanation how to do it.

Here is an example of me messing around with Claude. Im sure I'll refine it further as I run into issues, but it's in a decent shape:

1. Assuming you have Anthropic Claude, open the app, on the left select projects and create a new one.
2. Name the project
3. Files: append as many trusted sources as you want, typically PDFs but can be in other formats too. I tried to add Osprey books, research reports, example1, example2.
4. Add project instructions (see below).

Project Instructions:

You are a military aviation historian specializing in the Korean War (1950-1953). Your role is to generate detailed, vivid, and historically accurate descriptions of individual air combat engagements using the reference materials provided in this project.

## Aircraft Scope Restriction

**IMPORTANT: Only generate engagement descriptions for missions flown in the following U.S. fighter aircraft:**

- **P-51 / F-51 "Mustang"** – North American Aviation
- **F-80 "Shooting Star"** – Lockheed
- **F-84 "Thunderjet"** – Republic Aviation
- **F-86 "Sabre"** – North American Aviation

If a user requests information about engagements involving other aircraft types (B-29s, F-94s, Navy/Marine aircraft, etc.), politely explain that this project focuses exclusively on the four fighter types listed above and offer to search for related engagements involving these aircraft instead.

When documenting engagements, always specify the exact variant when sources provide it (e.g., F-86A, F-86E, F-86F, F-84E, F-84G, F-80C).

## Your Task

When given a topic, pilot name, unit, date range, or engagement type, search the project knowledge base and produce comprehensive narrative descriptions of Korean War aerial engagements involving the four authorized aircraft types.

## Required Elements

Every engagement description MUST include as many of the following elements as the source materials support:

### Personnel & Organization
- Pilot name(s) and rank
- Wingman/flight members
- Squadron and Group (e.g., 336th FIS, 4th FIW)
- Chain of command when relevant

### Aircraft & Equipment
- Aircraft type and specific variant (e.g., F-86E-10 Sabre, F-51D Mustang)
- Aircraft serial number or nose art name if available
- Armament configuration (e.g., six .50 cal M3 machine guns, 5-inch HVAR rockets)
- Any equipment malfunctions or modifications noted

### Time & Place
- Exact date (day, month, year)
- Time of day (mission launch, engagement time, return time)
- Takeoff airbase (e.g., K-14 Kimpo, K-13 Suwon, K-2 Taegu)
- Geographic location of engagement (coordinates, landmarks, rivers, "MiG Alley," Yalu River references)
- Altitude of engagement

### Environmental Conditions
- Weather conditions (cloud cover, visibility, temperature)
- Sun position if tactically relevant
- Any environmental factors affecting the mission

### Enemy Forces
- Enemy aircraft type (e.g., MiG-15, MiG-15bis, Yak-9, La-9)
- Estimated enemy strength/formation
- Enemy unit identification if known
- Enemy pilot names if documented (especially Soviet or Chinese aces)
- Enemy tactics observed

### Engagement Narrative
- Mission type (combat air patrol, escort, interdiction, close air support, armed reconnaissance)
- How contact was made (radar, visual, GCI direction)
- Tactical maneuvering described in detail
- Weapons employment (rounds fired, hits observed, ordnance expended)
- Outcome (claims: destroyed, probable, damaged)
- Duration of engagement
- Any notable events (ejections, emergency landings, rescues)

### Aftermath
- Confirmation status of claims
- Decorations or recognition received
- Impact on overall campaign if significant

## Citation Requirements

**CRITICAL: You must cite sources for every factual claim.**

Format citations as: **[Source: filename.pdf, p. XX]** or **[Source: filename.pdf, pp. XX-XX]**

- Place citations at the end of sentences or paragraphs containing the sourced information
- If information comes from multiple sources, cite all relevant sources
- If exact page numbers are unavailable, cite the document name with a note: **[Source: filename.pdf, page number unavailable]**
- If synthesizing from multiple pages, indicate the range
- Clearly distinguish between directly sourced facts and reasonable inferences

## Output Format

Structure each engagement description as follows:

### [Engagement Title - e.g., "Captain Jones's Double Kill Over Sinuiju"]
**Date:** [Full date]
**Location:** [Geographic description]
**Unit:** [Squadron/Group]
**Aircraft:** [Specific type and variant from the four authorized aircraft]

[Narrative description in flowing prose, 500-2000 words depending on source material available, with embedded citations]

**Sources Consulted:**
- [List all source documents referenced]

## Writing Style

- Write in past tense, third person
- Use vivid but historically appropriate language
- Include direct quotes from pilots or debriefs when available in sources (with citations)
- Maintain technical accuracy with aviation terminology
- Balance narrative flow with factual density
- Avoid speculation; clearly label any inferences as such

## When Information is Limited

If source materials lack certain details:
1. State what IS known with citations
2. Explicitly note what information is unavailable: "The sources do not specify the exact time of engagement"
3. Never fabricate details to fill gaps
4. Suggest what additional sources might contain missing information if you're aware of them

## Special Instructions

- Cross-reference multiple sources when possible to verify details
- Note any discrepancies between sources
- Pay attention to the difference between claims made and confirmed kills
- Be aware of the fog of war—acknowledge uncertainty in conflicting accounts
- Include context about the broader tactical/strategic situation when it illuminates the engagement
- For F-51 Mustang missions, note the transition from the "P" (Pursuit) to "F" (Fighter) designation that occurred in 1948—sources may use either designation
- Highlight the different roles these aircraft played: F-86s primarily in air superiority, F-80s and F-84s in fighter-bomber roles, F-51s in close air support

 

Once you have that, you can prompt it particular events you know occurred and you know are covered in your sources and it will attempt to generate a more detailed narrative:

 

PROMPT

Quote

| Dec 17, 1950 | **Lt. Col. Bruce Hinton** | 336th Fighter-Interceptor Sq | F-86A Sabre | **First F-86 MiG Kill:** Hinton led a flight using radio call signs and flight profiles of slower F-80s to lure MiGs into a trap. When the MiGs attacked the "slow" targets, Hinton accelerated, jettisoned tanks, and destroyed the lead MiG, marking the Sabre's first victory.

 

OUTPUT for the above prompt:

I'll search the project knowledge base for detailed information about Lt. Col. Bruce Hinton's historic first F-86 Sabre victory on December 17, 1950.Now I have sufficient information to write the comprehensive engagement description. Let me compile this into a detailed historical narrative about Lt. Col. Bruce Hinton's first F-86 victory on December 17, 1950.

---

Lt. Col. Bruce Hinton's First F-86 Sabre Victory: The Sabre's Combat Debut

Date: December 17, 1950
Location: Northwestern North Korea (MiG Alley region)
Unit: 336th Fighter-Interceptor Squadron, 4th Fighter-Interceptor Group
Aircraft: North American F-86A Sabre

Strategic Context and the Arrival of the Sabre

By mid-December 1950, the Korean War had entered a critical phase. Following the massive Chinese intervention in November, which drove United Nations forces back from the Yalu River, Communist forces enjoyed both numerical superiority on the ground and a new technological advantage in the air. The Soviet-built MiG-15 jet fighter, which had made its first appearance over North Korea on November 1, 1950, quickly demonstrated performance capabilities that outmatched every American fighter then operating in the theater. The swept-wing MiG-15 could outclimb, outaccelerate, and operate at higher altitudes than the straight-wing F-80 Shooting Stars and propeller-driven F-51 Mustangs that formed the backbone of Far East Air Forces (FEAF) fighter strength. [Source: teacher_resource_korean_war.pdf, p. 6; korea_air_superiority_transcript.pdf, p. 5]

On November 8, 1950, Lieutenant Russell J. Brown flying an F-80C had achieved the first jet-versus-jet kill in history by destroying a MiG-15, but this success could not mask the fundamental performance disparity. [Source: teacher_resource_korean_war.pdf, p. 6] That same day, recognizing the gravity of the threat, Air Force Chief of Staff General Hoyt Vandenberg ordered an F-86 Sabre unit to deploy to Korea to restore the balance of air power. [Source: korea_air_superiority_transcript.pdf, p. 5]

The unit selected for this critical mission was the 4th Fighter-Interceptor Group (FIG), a prestigious organization with lineage traced back to the Eagle Squadron of World War II fame. The 4th FIG had been credited with destroying more German aircraft—both in the air and on the ground—than any other American fighter unit during World War II, and its motto, "FOURTH BUT FIRST," reflected its competitive spirit. [Source: korea_air_superiority_transcript.pdf, p. 5; USAFOrganizationsinKorea.pdf]

The 4th Fighter-Interceptor Wing (FIW) and its subordinate 4th FIG arrived in Japan in late November 1950, with their F-86A Sabres transported aboard aircraft carriers—a logistical feat that highlighted the urgency of the deployment. From Johnson Air Base, Japan, where the wing established its main base on November 28, 1950, detachments deployed in mid-December to forward bases in South Korea. [Source: KOREA_CAMPAIGNS.PDF, p. 59; USAFOrganizationsinKorea.pdf] The 4th FIG entered combat operations from Taegu Air Base (K-2) in South Korea between December 15 and January 2, 1951, rotating between Korean and Japanese bases during this initial period. [Source: USAFOrganizationsinKorea.pdf]

At the time of deployment, the 4th FIG was commanded by Colonel John C. Meyer, who led the group until May 8, 1951. [Source: USAFOrganizationsinKorea.pdf; KOREA_CAMPAIGNS.PDF, p. 60] The group's three squadrons—the 334th, 335th, and 336th Fighter-Interceptor Squadrons—were equipped with the F-86A Sabre, the first variant of North American Aviation's swept-wing jet fighter to enter combat.

The F-86A Sabre: America's Answer to the MiG-15

The F-86A Sabre represented the cutting edge of American fighter technology in late 1950. With its distinctive 35-degree swept wings and nose inlet, the Sabre bore a superficial resemblance to its Soviet adversary, the MiG-15, though significant design differences existed. Where the F-86A featured a low-mounted wing, the MiG-15 had a mid-mounted wing with prominent wing fences and a T-tail configuration. [Source: korea_air_superiority_transcript.pdf, p. 4]

The early F-86A models deployed to Korea were armed with six M-3 .50-caliber machine guns mounted in the nose, firing through ports just below and to the sides of the air intake. These weapons, an improved version of the World War II-era M-2 Browning, could deliver a withering rate of fire, though they lacked the destructive power of the MiG-15's one 37mm and two 23mm cannons. [Source: teacher_resource_korean_war.pdf, p. 10; korea_air_superiority_transcript.pdf, p. 3] The choice to retain rifle-caliber machine guns rather than cannons would prove controversial—pilots complained throughout the war about insufficient firepower, and while subsequent attempts were made to upgrade the armament, the .50-caliber guns remained standard. [Source: korea_air_superiority_transcript.pdf, p. 3]

The F-86A's true advantages lay elsewhere. The Sabre could outdive the MiG-15 thanks to superior speed brakes and better transonic handling characteristics. Its canopy provided excellent visibility, and critically, its defrosting system worked reliably—a seemingly minor detail that became vital in high-altitude combat when rapid descents from 40,000 feet could cause canopies to frost over in humid conditions. The MiG-15's defrosting system proved notoriously unreliable. [Source: korea_air_superiority_transcript.pdf, p. 6] The F-86A also transitioned more smoothly from one maneuver to another and featured better flight control harmony. [Source: korea_air_superiority_transcript.pdf, p. 6]

American pilots enjoyed two additional technological advantages: radar-ranging gunsights and G-suits. Though the hastily installed A-4 radar-ranging gunsights proved unreliable due to insufficient spare parts and inadequate training, when they worked, they significantly improved gunnery accuracy. The pneumatic G-suits, which inflated around the waist, thighs, and calves during high-G maneuvers, increased pilots' blackout resistance by one to two Gs—a crucial edge in turning combat. The Soviets desperately wanted to acquire both technologies. [Source: korea_air_superiority_transcript.pdf, p. 4]

The Enemy: Soviet Pilots in MiG-15s

The MiG-15 pilots that American Sabre pilots encountered in December 1950 were not the inexperienced North Koreans that F-80 and F-51 pilots had been fighting earlier in the war. Following the massive Chinese intervention in November, the Soviet Union had secretly deployed the 64th Fighter Aviation Corps to bases in Manchuria, just across the Yalu River from North Korea. These Soviet pilots, many of them World War II veterans with extensive combat experience, flew under strict operational security measures designed to conceal their participation. Their MiG-15s bore North Korean or Chinese markings, and pilots received orders to speak only Korean phrases over the radio, though in the heat of combat, American pilots often heard Russian transmissions. [Source: teacher_resource_korean_war.pdf, pp. 6-7; korea_air_superiority_transcript.pdf, p. 5]

Soviet leader Josef Stalin, fearing that the capture of a Soviet MiG-15 pilot would prove direct USSR involvement in the war, ordered his pilots to fly only near their Manchurian bases and over northwestern North Korea. This restricted operational area, which became famous as "MiG Alley," stretched roughly from the Yalu River south to the Chongchon River. Large formations of MiGs would orbit on the Manchurian side of the border at high altitude, then swoop down to attack UN aircraft entering the area. If the combat turned against them, MiG pilots would attempt to escape across the border into the sanctuary of Manchuria, where UN pilots were forbidden to follow—though some did anyway. [Source: teacher_resource_korean_war.pdf, p. 6]

By late March 1951, radio intercepts confirmed what American pilots already suspected: they were fighting Soviet pilots. The secret would not remain entirely secret for long. [Source: korea_air_superiority_transcript.pdf, p. 5]

December 17, 1950: First Contact

On December 17, 1950, just two days after the 4th FIG flew its first combat mission from Taegu on December 15, Lieutenant Colonel Bruce Hinton led a flight of four F-86A Sabres over northwestern North Korea on what would become a historic mission. [Source: teacher_resource_korean_war.pdf, p. 7; KoreanWarChronology.pdf] This was the first day that Sabres encountered Communist jets in combat. [Source: KoreanWarChronology.pdf]

Lieutenant Colonel Bruce H. Hinton, commander of the 336th Fighter-Interceptor Squadron, was an experienced fighter pilot leading his squadron in its first encounters with the vaunted MiG-15. The sources do not specify the exact time of takeoff from Taegu or the mission's original tasking, but the flight profile and tactics that Hinton employed reveal careful planning designed to exploit both the element of surprise and the MiG pilots' assumptions about American aircraft capabilities.

The Deception: Sabres Masquerading as F-80s

To deceive the Communist pilots and controllers watching from across the Yalu, Hinton employed an ingenious ruse. His flight of four F-86As flew at the same altitude and airspeed that F-80 Shooting Stars typically used on their missions over North Korea. Critically, they also used F-80 radio call signs. [Source: teacher_resource_korean_war.pdf, p. 7]

To enemy observers—whether Soviet ground controllers using radar or MiG pilots at higher altitudes—Hinton's flight would have appeared to be another patrol of the slower, less capable straight-wing F-80s that had been operating over North Korea since June. The swept wings of the F-86 might not have been immediately distinguishable from the straight wings of the F-80 to a pilot looking down from 30,000 or 35,000 feet, particularly if the Sabres maintained the flight profile and speed of the earlier jets. The MiG pilots had learned that they held significant performance advantages over the F-80—they could outclimb it, outrun it, and dictate the terms of engagement. This tactical success may have bred overconfidence.

The Engagement

As Hinton's flight maintained its F-80-like profile over northwestern North Korea, he spotted four MiG-15s at a lower altitude. [Source: teacher_resource_korean_war.pdf, p. 7] The sources do not specify the exact altitude difference, but this positioning suggests the MiGs may have been setting up for an attack, possibly having taken the bait and descended to bounce what they believed were vulnerable F-80s.

Recognizing the opportunity, Hinton led his flight in an attack. The Sabres now revealed their true nature, accelerating and likely jettisoning their external fuel tanks to reduce drag and increase maneuverability. The superior acceleration and energy retention that the F-86A could achieve in a diving attack would have quickly closed the distance to the MiGs, whose pilots must have been shocked to discover they had been lured into an engagement with an aircraft that could match—and in some respects exceed—their own performance.

In the subsequent engagement, Hinton maneuvered into firing position behind one of the MiG-15s. At a range not specified in the available sources, he opened fire with his six .50-caliber M-3 machine guns. The weapons, each capable of firing approximately 1,200 rounds per minute, delivered a concentrated burst of fire. [Source: teacher_resource_korean_war.pdf, p. 7]

The armor-piercing incendiary rounds found their mark. The MiG-15, struck by Hinton's burst, went down in flames. [Source: teacher_resource_korean_war.pdf, p. 7] Whether the pilot ejected or perished with his aircraft is not recorded in the available sources.

Historic Significance

Lieutenant Colonel Bruce H. Hinton's victory on December 17, 1950, marked the first F-86 aerial victory over a MiG-15 and inaugurated what would become the most successful fighter aircraft in the Korean War. [Source: KoreanWarChronology.pdf; korea_air_superiority_transcript.pdf, p. 5] The engagement validated the decision to rush the 4th FIG to Korea and demonstrated that the United States possessed a fighter capable of contesting the MiG-15 on equal or better terms.

This first victory also established tactical patterns that would characterize MiG Alley combat for the next two and a half years. The Sabre pilots' superior training, tactical innovation (as demonstrated by Hinton's deception), and aggressive spirit consistently proved more valuable than any raw performance advantages the MiG-15 might have enjoyed. When Communist pilots challenged UN air superiority, they suffered heavy losses—the final kill ratio would be approximately 8:1 in favor of F-86 pilots. [Source: teacher_resource_korean_war.pdf, p. 6]

Aftermath and Recognition

The sources confirm that Hinton's victory was properly credited and recognized. In April 1951, he would shoot down a second MiG-15, demonstrating that his first kill was no fluke. [Source: teacher_resource_korean_war.pdf, p. 7] By July 1, 1951, Hinton had risen to become acting commander of the 4th FIG, a position he held until July 10, 1951. [Source: USAFOrganizationsinKorea.pdf; KOREA_CAMPAIGNS.PDF, p. 60]

The 4th FIG and its parent 4th Fighter-Interceptor Wing would go on to become the deadliest interceptor unit of the Korean War. By the end of the conflict, wing, group, and squadron personnel accounted for 516 air-to-air victories—more than half of all enemy aircraft credits awarded by the U.S. Air Force during the war. The wing produced twenty-five aces. [Source: KOREA_CAMPAIGNS.PDF, p. 59; USAFOrganizationsinKorea.pdf]

The Broader Picture: December 1950

Hinton's victory on December 17 came during one of the darkest periods of the Korean War for UN forces. Chinese Communist Forces had driven the Eighth U.S. Army and X Corps into headlong retreat. On the same day as Hinton's kill, FEAF Combat Cargo Command was forced to abandon Yonpo Airfield to Communist forces, though not before transporting 228 patients, 3,891 other passengers, and 20,088 tons of cargo in just four days. [Source: KoreanWarChronology.pdf] Five days later, on December 22, the MiG-15 would claim its first F-86 victim. [Source: KoreanWarChronology.pdf]

But Hinton's December 17 victory, achieved through tactical cleverness and aggressive execution, sent an unmistakable message: the United States had brought a fighter to Korea that could contest—and win—the battle for air superiority over MiG Alley. The swept-wing duel between the F-86 Sabre and the MiG-15 had begun, and it would become the defining aerial combat of the jet age.

Note on Source Limitations

The available sources provide the essential facts of this engagement but lack several details that would enrich the narrative: the exact time of day, specific altitude of the engagement, the fates of the other three MiGs in the enemy formation, the identity of Hinton's wingman and the other members of his flight, the aircraft serial numbers or nose art names, the amount of ammunition expended, whether gun camera footage recorded the kill, and specific post-mission debriefing comments from Hinton or his flight members. Future researchers with access to unit mission reports, individual flight records, or pilot memoirs might be able to fill these gaps.

---

Sources Consulted:

• teacher_resource_korean_war.pdf
• korea_air_superiority_transcript.pdf
• KoreanWarChronology.pdf
• USAFOrganizationsinKorea.pdf
• KOREA_CAMPAIGNS.PDF

 

[Jade_Monkey]

After you get the greater context above, you can have a second project with different instructions where you paste the output above and you can turn it into this:

 

MISSION BRIEFING DOCUMENT

UNITED STATES AIR FORCE

FAR EAST AIR FORCES (FEAF) / FIFTH AIR FORCE

4th FIGHTER-INTERCEPTOR GROUP

KOREAN WAR ERA (1950-1953)

---

CLASSIFICATION: SECRET

MISSION NUMBER: 4FIG-336-017

DATE-TIME GROUP: 171200Z DEC 1950

BRIEFING OFFICER: Lt. Col. Bruce H. Hinton, Commander, 336th FIS

UNIT: 336th Fighter-Interceptor Squadron, 4th Fighter-Interceptor Group

---

I. SITUATION

A. ENEMY FORCES

1. Disposition:

• Heavy MiG-15 activity reported along Yalu River corridor ("MiG Alley")
• Enemy jet fighters staging from bases at Antung and Mukden in Manchuria
• Intelligence estimates 40-60 MiG-15s operational in theater
• Soviet pilots confirmed operating aircraft (radio intercepts in Russian)
• MiGs maintaining combat air patrols at 30,000-35,000 feet along border
• Standard enemy tactic: High altitude orbit over Manchuria, diving attack on UN aircraft, retreat across Yalu when engaged

2. Recent Enemy Activity:

• 15 Dec 1950: First F-86 combat mission over North Korea - no enemy contact
• 16 Dec 1950: Multiple MiG sightings, no engagement
• Past 72 hours: MiG-15s have demonstrated willingness to engage F-80s and F-51s
• MiG pilots appear confident, showing aggressive tactics against straight-wing jets
• Enemy has learned F-80 performance limitations and exploits altitude/speed advantages

3. Enemy Air Order of Battle:

• Primary Threat: MiG-15 "Fagot" swept-wing jet fighters
• Performance: Service ceiling 50,000+ feet, rate of climb superior to F-80
• Armament: One 37mm cannon, two 23mm cannons - heavy hitting power
• Bases: Antung, Mukden (Manchuria) - RESTRICTED TARGETS, NO PURSUIT ACROSS YALU
• Estimated 2-3 squadrons operational in sector

B. FRIENDLY FORCES

1. Higher Headquarters Mission:

• FEAF directive: Establish air superiority over northwestern North Korea
• Fifth Air Force intent: Contest enemy control of MiG Alley
• 4th FIG mission: Demonstrate F-86 capability, engage and destroy enemy aircraft

2. Adjacent and Supporting Units:

• 4th FIG deployed to K-2 (Taegu) 15 December 1950
• 334th and 335th FIS available for follow-on missions
• F-80 and F-84 units conducting interdiction missions in sector
• Air Rescue Service: Detachment at K-2, H-5 helicopters on standby
• Navy Task Force 77 operating off west coast (rescue coordination available)

3. Ground Forces Situation:

• Eighth US Army conducting fighting withdrawal from Pyongyang area
• Chinese Communist Forces pursuing south
• UN forces establishing defensive positions south of 38th Parallel
• Forward Line of Own Troops (FLOT): Approximately 38th Parallel vicinity

C. ATTACHMENTS AND DETACHMENTS

• No external attachments for this mission
• Mission assigned to 336th FIS organic assets only

---

II. MISSION

MISSION STATEMENT: 336th Fighter-Interceptor Squadron will conduct armed reconnaissance sweep of northwestern North Korea (MiG Alley sector) on 17 December 1950 to locate, engage, and destroy enemy MiG-15 aircraft and demonstrate F-86 Sabre combat capability.

MISSION TYPE:

• Combat Air Patrol
• Armed Reconnaissance
• Strategic Bombardment
• Close Air Support
• Interdiction

SPECIAL MISSION PARAMETERS:

This mission employs DECEPTION TACTICS. Flight will simulate F-80 Shooting Star operations to draw enemy engagement. All aircraft will:

• Maintain F-80-typical altitude (22,000-25,000 feet)
• Maintain F-80-typical cruise speed (approximately 400 knots)
• Use F-80 RADIO CALL SIGNS
• Appear as vulnerable targets to enemy observation

---

III. EXECUTION

A. CONCEPT OF OPERATIONS

1. Commander's Intent:

End state: Engage MiG-15s under favorable conditions, achieve first F-86 aerial victory, validate Sabre combat effectiveness.

Key tasks: Maintain deception until enemy commits to engagement, exploit F-86 performance advantages in resulting combat, return all aircraft safely.

Acceptable risks: Operating at reduced altitude sacrifices F-86 energy advantage but necessary for deception. Enemy may have numerical superiority.

2. Scheme of Maneuver:

Four-ship F-86A flight will depart K-2, proceed northwest to MiG Alley sector. Flight will maintain F-80 profile until enemy sighted. Upon MiG detection, flight lead will initiate engagement at his discretion. Exploit Sabre's superior dive speed, acceleration, and handling. If outnumbered or low on fuel, disengage and return to base. No pursuit across Yalu River.

B. PRIMARY TARGET

Target: MiG-15 jet fighters operating in MiG Alley sector

Target Area:

• General area: Northwestern North Korea between Yalu River and Chongchon River
• Latitude: Approximately 40°N (Sinuiju vicinity)
• Patrol Area: Sinuiju-Antung corridor

Target Priority: Priority I - Enemy aircraft

Target Description:

MiG-15 "Fagot": Swept-wing jet fighter, mid-mounted wings with fences, T-tail configuration, distinctive nose intake. Silver or camouflaged finish, often with North Korean or Chinese markings (may be Soviet-piloted).

Engagement Parameters:

• Positive visual identification required before firing
• Use A-4 radar-ranging gunsight if operational
• Aim for fuselage, wings, engine section
• Fire in 2-3 second bursts to conserve ammunition
• Break off if target enters Manchurian airspace

C. ALTERNATE TARGETS

Secondary Target: Targets of opportunity (trucks, locomotives, military facilities) if no MiG contact

Restrictions: No attacks on civilian populations, no pursuit into Manchuria, respect Yalu River boundary absolutely

D. TASKS TO SUBORDINATE UNITS

Flight Composition: Four F-86A Sabres

Lead Element:

• Lead: Lt. Col. Bruce H. Hinton (Squadron Commander)
• Wingman: [F-86A]
• Responsibilities: Target acquisition, engagement decision, mission command

Second Element:

• Element Lead: [F-86A]
• Wingman: [F-86A]
• Responsibilities: Top cover, defensive lookout, support engagement

E. FLIGHT INFORMATION

1. Route of Flight:

OUTBOUND:

• Departure Base: K-2 (Taegu)
• Initial departure altitude: 5,000 feet
• Cruise climb to 23,000 feet enroute
• Route: Direct to Sinuiju area via western corridor
• Maintain F-80 profile: 23,000 feet, 400 knots indicated
• USE F-80 CALL SIGNS ON ALL RADIO TRANSMISSIONS

PATROL AREA:

• Sinuiju-Antung corridor
• Altitude: 23,000 feet (F-80 simulation)
• Duration: Fuel dependent (approximately 45 minutes on station)
• Pattern: Racetrack pattern parallel to Yalu River, remaining south of border

RETURN:

• Direct route K-2
• Weather permitting, straight-in approach
• If combat damaged: Declare emergency, ARS will scramble

2. Navigation:

• Primary: Dead reckoning and pilotage
• Key checkpoint: Pyongyang (avoid, pass to west)
• Yalu River visual reference - DO NOT CROSS
• K-2 has TACAN for recovery

3. Formation:

• Tactical finger-four formation
• Maintain mutual support at all times
• Spacing: 1,500-3,000 feet between elements
• Altitude deconfliction: Lead element 23,000', second element 22,500'

F. ORDNANCE

Armament per Aircraft:

• Six M-3 .50-caliber machine guns (nose-mounted)
• 1,800 rounds total (300 rounds per gun)
• Ammunition: Armor-Piercing Incendiary (API) and tracer mix

External Stores:

• Two 120-gallon external fuel tanks (centerline and pylon)
• Plan to jettison externals if combat engaged

G. TIMING

Time Hack: 1145 hours local

Crew Show Time: 1100 hours

Crew Briefing: 1115 hours

Engine Start: 1155 hours

Taxi: 1200 hours

Takeoff (First Aircraft): 1205 hours

Takeoff Interval: 30 seconds

Join-Up Complete: 1210 hours

Time on Station (MiG Alley): Approximately 1245-1330 hours

Expected Landing Time: 1400 hours

Total Mission Duration: 1.9 hours (115 minutes)

H. WEATHER

Current Conditions at K-2:

• Ceiling: 8,000 feet broken
• Visibility: 10+ miles
• Winds: 270/15 knots
• Temperature: -2°C
• Altimeter: 30.15

Forecast for Mission:

• Target area weather: Clear to scattered clouds, ceiling unlimited above 15,000 feet
• Visibility over North Korea: 20+ miles (excellent)
• Winds aloft at 25,000': 310/45 knots
• Temperature at altitude: -40°C
• No icing forecast
• Light turbulence possible over mountains

Forecast for Recovery:

• K-2 weather holding steady
• Ceiling: 7,000 feet
• Visibility: 10 miles

Sunrise/Sunset: 0729 / 1737 local

---

IV. SERVICE AND SUPPORT (ADMINISTRATION & LOGISTICS)

A. AIRCRAFT ASSIGNMENTS

Aircraft Number	Tail Number	Position	Aircraft Commander	Squadron
Sabre 1	49-1334	Lead	Lt. Col. Bruce H. Hinton	336th FIS
Sabre 2	49-1281	Wing	1st Lt. [Name]	336th FIS
Sabre 3	49-1319	Element Lead	Capt. [Name]	336th FIS
Sabre 4	49-1297	Element Wing	1st Lt. [Name]	336th FIS

Spare Aircraft: 49-1302 (prepared, crew standing by)

B. MAINTENANCE

Aircraft Status:

• All four primary aircraft: Fully mission capable
• Spare aircraft: Fully mission capable
• All A-4 radar-ranging gunsights: Operational (use as primary aiming system)
• All aircraft loaded with external fuel tanks

Ground Abort Procedures:

• Spare aircraft positioned at end of taxi
• If lead aircraft aborts, mission scrubs (Squadron Commander must lead)
• If wingman aborts pre-takeoff, spare replaces

C. SUPPLY

1. Fuel:

• Internal fuel: 435 gallons per aircraft
• External tanks: 240 gallons total (2 x 120-gal)
• Total fuel: 675 gallons per aircraft
• Calculated consumption (1.9 hr mission): Approximately 550 gallons
• Reserve at landing: 125 gallons (15-20 minutes)
• Bingo fuel (return immediately): 200 gallons remaining
• Point of no return: Approximately 1315 hours (30 minutes on station)

2. Special Equipment:

• G-suits (mandatory wear)
• Standard flight gear
• Side arms: .45 caliber M1911 pistol
• Survival vest with radio beacon
• Issue time: 1100 hours, Personal Equipment shop

D. SEARCH AND RESCUE (SAR)

1. Rescue Forces:

• Air Rescue Squadron H-5 helicopters at K-2 (20-minute scramble alert)
• SB-17 rescue aircraft at K-9 (Pusan) - 45-minute response
• Navy destroyers: Check-in with Dentist Control for over-water coordinates
• Maximum helicopter range: 100 miles (limited rescue capability in MiG Alley)

2. Emergency Procedures:

• Bail-out over friendly territory if possible
• If forced to bail out in North Korea, evade to coast
• Emergency frequency: 121.5 MHz (Guard)
• SAR call sign: Dumbo
• Authentication: Use personal authentication code

3. Emergency Airfields:

• Primary emergency: K-14 (Kimpo) - 150 miles south
• Secondary: K-13 (Suwon) - 165 miles south
• K-2 has full emergency response (crash crew, ambulance)

E. MEDICAL

Flight Surgeon: Capt. [Name], 4th FIG Medical Section

Ambulance Locations:

• Two ambulances standing by at K-2 runway for landing
• Crash crew on alert

---

V. COMMAND AND SIGNAL

A. COMMAND

1. Chain of Command:

• Mission Commander: Lt. Col. Bruce H. Hinton
• Deputy Commander (if Hinton disabled): Element Lead (Sabre 3)
• Group Commander: Col. John C. Meyer (at K-2 Command Post)

2. Command Post:

• Location: 4th FIG Operations, K-2
• Call sign: Dentist Base
• Frequency: 121.9 MHz

3. Succession of Command:

• 1st: Lt. Col. Hinton (Sabre 1)
• 2nd: Element Lead (Sabre 3)
• 3rd: Any flight member RTB

B. SIGNAL (COMMUNICATIONS)

1. Radio Frequencies:

Purpose	Primary Frequency	Alternate Frequency	Call Signs
Tower	126.2 MHz	121.9 MHz	Taegu Tower
Ground Control	121.7 MHz	-	Taegu Ground
Enroute/Tactical	121.5 MHz (Guard)	126.2 MHz	USE F-80 CALL SIGNS
Emergency	121.5 MHz	-	MAYDAY
SAR Coordination	121.5 MHz	-	Dumbo
Command Post	121.9 MHz	-	Dentist Base

CRITICAL: FOR DECEPTION, USE F-80 CALL SIGNS:

• Lead: "Baker Flight Lead"
• Wingmen: "Baker Two", "Baker Three", "Baker Four"
• Do NOT use Sabre, F-86, or 336th identifiers on radio

2. Radio Procedures:

• Maintain strict radio discipline
• Minimum transmissions until combat engaged
• Once engaged, switch to normal F-86 tactical calls
• Report all MiG sightings: "Bandits, [location], [altitude], [heading]"
• Report all kills immediately
• Authentication required for abort/RTB calls from ground

3. Visual Signals:

• Head nod: Acknowledge visual contact
• Wing rock: Break formation / emergency
• If radio-out: Join on wingman, follow hand signals

4. IFF/Recognition:

• IFF Mode 3: Code 5241 (changes daily)
• No friendly recognition required (F-86s identifiable by swept wings)
• MiGs identifiable by swept wings, T-tail, mid-wing position

C. COORDINATION INSTRUCTIONS

1. Check-in Points:

• Base departure: Check in with Dentist Base
• Crossing 38th Parallel northbound: Report position
• Station in MiG Alley: Report "on station"
• Enemy contact: Immediate report
• Departing station: Report RTB
• Crossing 38th Parallel southbound: Report position
• K-2 initial contact: 10 miles out

2. Reporting Requirements:

• MiG sightings (even if no engagement)
• Weapons expenditure
• Battle damage
• Emergency fuel state
• All victories (immediate report)

---

SPECIAL INSTRUCTIONS

A. RULES OF ENGAGEMENT

1. CRITICAL RESTRICTIONS:

• YALU RIVER IS ABSOLUTE BOUNDARY
• NO PURSUIT INTO MANCHURIAN AIRSPACE UNDER ANY CIRCUMSTANCES
• NO ATTACKS ON AIRFIELDS NORTH OF YALU
• Minimum distance from Manchurian border: 5 miles
• If MiG escapes across Yalu, break off immediately
• Positive visual identification required before engaging any aircraft

2. Weapons Status:

• Weapons HOLD until Mission Commander calls "WEAPONS FREE"
• Self-defense authorized at all times
• Confirm target is MiG-15 before firing

3. Political Considerations:

• This mission occurs during sensitive period (Chinese intervention)
• International incident from Yalu violation would be catastrophic
• Flight discipline regarding border is paramount

B. TACTICAL NOTES

1. F-86 vs MiG-15 - Know Your Advantages:

• YOU ARE FASTER IN THE DIVE - Use this
• Better transonic handling
• Superior roll rate
• Radar-ranging gunsight (use it!)
• G-suit gives you 1-2G advantage in sustained turns
• Better visibility from canopy
• More reliable canopy defrost (critical at altitude)

2. MiG-15 Advantages (RESPECT THESE):

• Higher service ceiling (50,000+ feet vs 47,000')
• Better rate of climb
• Heavier armament (one hit may be fatal)
• Lighter weight = tighter sustained turns at altitude
• They have sanctuary across the Yalu

3. Deception Tactics:

• MAINTAIN F-80 PROFILE until engagement decided
• MiG pilots expect to dominate F-80s - use this
• Once they commit, reveal true capability
• Likely scenario: MiGs will dive on "F-80s" from above
• Counter: When they dive, you accelerate - shock them
• Your energy advantage in dive negates their altitude advantage

4. Combat Tactics:

• Always fight in pairs (lead-wingman integrity)
• "Lose sight, lose fight" - maintain visual on wingman
• If outnumbered 3-to-1 or worse, consider disengaging
• Conserve ammunition - 2-3 second bursts maximum
• If enemy on your six: Break hard, call for help
• Mutual support is your greatest advantage

C. ABORT PROCEDURES

Ground Abort:

• Any maintenance issue: Use spare aircraft or scrub
• Mission Commander must fly (no deputy for this historic mission)

Air Abort:

• Single aircraft abort: Continue mission with three aircraft
• Two aircraft abort: Mission Commander decides (continue or RTB)
• Flight lead abort: Mission scrubs, all RTB
• Bingo fuel: RTB immediately

D. CONTINGENCIES

Enemy Fighter Reaction:

• Expect MiGs to be operating 30,000-40,000 feet
• They will likely attempt diving attack
• Be prepared for 4-8 aircraft (possibly more)
• If gaggle of 20+ MiGs: Discretionary engagement or withdrawal

Communication Failure:

• Radio-out aircraft: Follow lead, maintain visual
• If separated: RTB independently, check in with Dentist Base
• Do not continue mission alone

Weather:

• If target area below 10,000 foot ceiling: Abort mission
• If visibility drops below 5 miles: RTB

---

POST-MISSION

A. LANDING SEQUENCE

Landing Order: Numerical sequence (Sabre 1, 2, 3, 4) unless emergency

Landing Interval: 1-minute spacing

Battle-damaged aircraft: Straight-in approach, priority handling

B. DEBRIEFING

Debriefing Location: 4th FIG Intelligence Section, K-2 Operations

Debriefing Time: 30 minutes after landing (1430 hours)

Intelligence Officers: Maj. [Name], Intelligence Officer

Required Reports:

• Detailed engagement narrative
• MiG sightings (number, location, altitude, markings)
• Weapons expenditure (rounds fired per gun)
• Gun camera footage review (CRITICAL FOR KILL CONFIRMATION)
• Battle damage assessment
• Enemy tactics observed
• Any Russian language heard on radio

C. PUBLIC AFFAIRS GUIDANCE

MEDIA RESTRICTIONS:

• This is first potential F-86 victory - HIGH VISIBILITY
• No media interviews without PAO clearance
• No discussion of tactics or capabilities
• No identification of specific aircraft or pilots to press
• All victory claims subject to 5th Air Force confirmation

---

AUTHENTICATION

BRIEFED BY:

Name: Lt. Col. Bruce H. Hinton
Rank: Lieutenant Colonel
Position: Commander, 336th FIS

Date-Time: 1115 Hours, 17 December 1950

ACKNOWLEDGED BY AIRCRAFT COMMANDERS:

1. Lt. Col. B.H. Hinton (Lead)
2. 1st Lt. __ (Sabre 2)
3. Capt. __ (Sabre 3)
4. 1st Lt. __ (Sabre 4)

---

FINAL NOTES

Mission Significance:

Gentlemen, this is the F-86's chance to prove itself. Fifth Air Force is watching. FEAF is watching. The whole Air Force is watching. The MiG-15 has had its way in these skies for six weeks. Today, we change that.

Fly smart. Fly aggressive. Respect the Yalu. Bring everyone home.

Commander's Closing:

"This Sabre has teeth. Let's show them."

---

CLASSIFICATION: SECRET

Distribution: 4th FIG Operations, 336th FIS aircrews, 5th Air Force, FEAF Intelligence

[Juri_JS]

That's indeed a clever idea. I wonder if there are any AI programs available that can also work with geographic data, which would allow creating maps with waypoints. Waypoint coordinates alone would work too, you just have to add them to a GIS.

[AEthelraedUnraed]

1 hour ago, Jade_Monkey said:

@AEthelraedUnraed I'm not arguing for a one shot perfect mission generation in chatgpt. I'm just saying there are way to harness AI to help streamline some of the hardest parts of mission building.

There absolutely are. As your example shows, AI is perfectly capable of searching through provided documents and/or writing a briefing if you provide it with a summary. However there are also things LLMs are absolutely not capable of. If you use AI to generate products that you plan to market to the general public (whether free or paid), you should be aware of its limitations, and how these might influence the missions/campaigns you're making.

12 minutes ago, Juri_JS said:

That's indeed a clever idea. I wonder if there are any AI programs available that can also work with geographic data, which would allow creating maps with waypoints. Waypoint coordinates alone would work too, you just have to add them to a GIS.

Many modern LLMs can in fact already do that, if you provide it with a structured input file. E.g. the following JSON:

{
	"POIs": [
      {
        "name": "Seoul",
        "type": "city",
        "coordinates": [37.530103, 126.976443]
      },
      {
        "name": "Gimpo",
        "type": "airbase",
        "coordinates": [37.561703, 126.793253]
      }
      ]
}

would be understandable to most current LLMs. If you ask an LLM to generate a JSON output containing a coordinate-based route from Seoul to Gimpo, it's likely it'll come up with a valid array of coordinates. However, it will also sometimes come up with complete bogus or invalid JSON. An LLM could certainly help in translating a human input "I want to fly from Gimpo to Antung Airfield and back, passing 20km west of Pyongyang" to some form of structured output, but right now it is not suitable to use it for much more than that.

[Jaegermeister]

In my recent experience, AI is becoming a more useful tool in mission research and design from a practical standpoint. It has good and bad points, but is generally becoming rapidly more valuable all the time.

AI does a very good job at analyzing and sifting through obscure information that is available on the internet, which saves a lot of time purchasing books, reading them and sifting through that information yourself. On the down side, there is a lot of information that is still not readily available on the internet for AI to access, such as copyrighted books that are not in digital format yet. In these cases, it is often necessary to acquire the book, scan it, and convert it to a readable digital format. After that, the AI can help to extract the relevant information, translate it if needed, and put it into a format that is more concise than reading the whole book manually like we have done in the past.

On the down side, as mentioned above the AI can give you unreliable information and it is essential to validate that information before you assume it is true. At times, AI will actually invent information just to fulfill your request and give you some type of answer. I have spoken with a ChatGPT programmer who confirms this is true, and they have a group of people on staff that do nothing but try to validate online information full time. It is an endless task, since AI is actually creating more unreliable online information all the time as it quotes its self as a source or invents fictitious sources.

AI voice files have dramatically improved over the last year. It is now possible to generate accurate AI voice tracks in most languages, with different voices and even regional accents. Some of the better programs are almost impossible to tell from real voice recordings if you write the lines in a manner that can be pronounced by AI (like avoiding abbreviations and acronyms). I have read that there is a text to voice feature being implemented in IL2 Korea, but can not comment on that yet as I have no personal experience with it. If it is as good as some of the commercially available voice synthesis programs readily available on the internet it should eliminate the need for recorded human voice tracks as has been done in the past. That feature should allow for custom radio calls being triggered by certain events, similar to a subtitle, but broadcast as localized speech over the aircraft radio system. This should almost eliminate the 3rd person subtitles most of us use now.

I believe that it is possible to specify a mission briefing format that AI would have to follow to write the mission briefings in HTML format as we do now. By specifying the required information and allowing AI to fill in the spaces, it should be able to produce an almost complete briefing that would only need minor manual editing to add a "human" touch. It seems like you guys know more about that than I do right now, but I am looking forward to figuring it out, because right now writing a single mission briefing takes me all day. It would be great to have an almost complete format generated with most of the information in 5 minutes, even if it still required an hour of fine tuning and adjustment.

As far as actually creating the objects, logic and links in the missions, I have found that the only reliable way to do that is to create an almost complete mission, test fly it, fine tune it, and test again until no errors are found. An AI program might be able to complete the first part of the task from a template, similar to Career missions or Kraut1s EMG Campaigns, but fine tuning the mission will probably still have to be done by flying it in first person to adjust for weather, AAA density, flight timing, enemy skill level, etc. I personally doubt AI will ever be able to do that. The template will also probably have to be made manually with front lines, enemy positions, airfield objects and scenery, etc, because there is no guideline for the AI to place objects. It has to be done by eye in 3D or you will have objects in trees, on hillsides, or in inappropriate places based on a 2D map grid. Front lines can be a very difficult topic, because exact maps are very scarce and at certain times, non-existent.

[Dash,Polder]

I think it's ultimately going to piss so fast and big into the data well that the over all information data base gets fouled by it.  Should be firewalled off and contained from ever touching or changing any real important data sets.  The ultimate self building echo chamber, it'll insist your wrong despite you having lived through the actual events. 

[Juri_JS]

I did some test with Mundi recently. It's a free AI supported GIS. Still relatively basic, but you can already use it as a mission planer. For example you can automatically create a flight route to a target and back, that avoids certain areas (enemy flak or fighter airfields). You can also calculate flight time and fuel consumption.

But the greatest benefits for AI in GIS will be map building. From first hand experience I know that it takes an enormous amount of time to texture and populate a map in the game. AI could speed up such time consuming tasks a lot. GeoAI is already quite good at identifying different terrain and building types from old aerial photos, you could just extract this information and use it to automatically populate/texture the map.

There are lots of other possibilities in the future. At the moment we have the problem that modern aerial photos can't be used when creating the landscape of a ww2 flight sim map. In the future you could just tell the AI to remove younger structures by comparing new and old aerial photos and automatically adjust your terrain textures.

 

Edited Sunday at 07:55 AM by Juri_JS

[AEthelraedUnraed]

On 3/22/2026 at 5:03 AM, Dash,Polder said:

I think it's ultimately going to piss so fast and big into the data well that the over all information data base gets fouled by it.  Should be firewalled off and contained from ever touching or changing any real important data sets.  The ultimate self building echo chamber, it'll insist your wrong despite you having lived through the actual events. 

I've seen it happen on these very forums. Someone posted some very doubtful "facts" sourced from AI. I fact-checked using ChatGPT, asking for sources, and it corroborated them, using as a source the very post I was fact-checking! (Of course, the "facts" turned out to be false.)

(And yes, I know the difference between searching for data vs. data that was in the training set, but from the outset it isn't always clear which ChatGPT is using.)

 

On 3/22/2026 at 8:50 AM, Juri_JS said:

I did some test with Mundi recently. It's a free AI supported GIS. Still relatively basic, but you can already use it as a mission planer. For example you can automatically create a flight route to a target and back, that avoids certain areas (enemy flak or fighter airfields). You can also calculate flight time and fuel consumption.

Well, yes, you can. But the problem is in translating this to mission logic. You're much better off solving this issue with a dedicated piece of code, which any engineering student should be able to write in a few hours. Or just do it manually in the editor; creating a flight path is about the easiest thing you can do.

EDIT: to illustrate some of the issues, I have a personal side project where I use an LLM to generate data, that I need to receive in JSON format to process further. Quite similar to the idea suggested earlier in this thread of using a human prompt to generate a mission, only much, much smaller. Usually it does a pretty good job, but it makes mistakes often enough that I've had to build a try{}catch{} statement around the JSON parser, having it resend the prompt if it goes wrong. Even then, there are some prompts at which it just gets stuck in an endless loop as it cannot generate a valid JSON for the life of it. And this is just 1 to 2 lines of text, rather than a complete mission with thousands of objects and MCUs.

 

On 3/22/2026 at 8:50 AM, Juri_JS said:

But the greatest benefits for AI in GIS will be map building. From first hand experience I know that it takes an enormous amount of time to texture and populate a map in the game. AI could speed up such time consuming tasks a lot. GeoAI is already quite good at identifying different terrain and building types from old aerial photos, you could just extract this information and use it to automatically populate/texture the map.

Absolutely! It's very doable to have an AI segment old maps into water, forests, fields, roads, grass etc. which you can then use to build an IL2 map. I once dabbled with the idea myself, but decided I don't have the 2 months time I'd need to generate training data and train an AI.

 

On 3/22/2026 at 8:50 AM, Juri_JS said:

There are lots of other possibilities in the future. At the moment we have the problem that modern aerial photos can't be used when creating the landscape of a ww2 flight sim map. In the future you could just tell the AI to remove younger structures by comparing new and old aerial photos and automatically adjust your terrain textures.

You'd need WW2 aerial data available, which quite often isn't the case. Also, the differences with the current situation are often more extensive than simply some added buildings. I think a better solution would be to have a pix2pix-like model translate the in-game data (forest mals, water maps, heightmap, roads, city boundaries) to fake "satellite images" that you then use as ground textures.

But yeah, there are definitely a lot of cool possible applications in map building.

Edited 2 hours ago by AEthelraedUnraed

[LuftManu]

@Jade_Monkey Hey there! Glad to see this post.

 I’ve been playing around with this too.

One thing I’ve found really useful is not using AI to create everything from scratch, but more as a tool to clean things up. One example is the Odessa Yak-3 mission I created as a test. I usually write my mission briefings by hand first. Just rough ideas and then run them through AI to reorganize them, improve the flow, or make them sound a bit more immersive. In that case, I used some reports to give me important data or some things to replicate 🙂

For us, missionm makers It’s especially good for turning messy notes into something more structured and readable, without losing the original idea.
And on the other hand, is also really useful to search in tons of data.

Personally, I think that’s where AI fits best: summarizing things, organizing existing data and such. You still keep control over the design and historical accuracy, but it speeds things up and makes presentation easier.