progress update Charles Jameson progress update Charles Jameson

DESIGN,DEVELOP,REPEAT

The constellation slid free from its position in the sky, and took flight.

Fire In The Sky

In the past week we had an addition to our team. With over 10 years experience in the industry and a full-stack designer, Franz comes prepared and has already contributed to our overall design and provided essential industry guidance in product development and deployment. The insights of which have shaped our approach to this month of October and the work we have planned for Radiant Assistant.

Very difficult to get a dragon constellation that doesn’t turn into a Power Rangers character.

Franz joins us at an especially critical time as we’ve begun development of our most ambitious new addition to how RA will assist its players. This has lead to whole new sections of the website needing built. Which means, lots and lots of research, discovery, feedback, and a loop of that, until we’ve drilled down to something we hope our users will genuinely love.

In other news soon to deeply impact the AI generative world is the recently announced Dall-E 3. Another significant step forward in the industry and if you look at its art results compared side by side with its progenitor Dall-E 2, given the same prompt, it achieves markedly superior results. Along with an increase in image quality comes a tool-box of other improvements certain to shift what was thought possible in the field only a short while ago.

Image Credit: OpenAI

Dall-E 3 and doubtlessley the art generators that will follow it, will now be capable of creating impressively nuanced and tonally accurate images with only conversational input. The user need not be especially specific in their phrasing and the AI can infer what it should create. “I don’t know. Give me a happy hedgehog frolicking”. That’s a joke but it was similar language that created the image below.

Image Credit: OpenAI

Objects in relation to each other is a concept the AI now grasps far better. Tell it that you want grass, adjacent to a box, adjacent to a telephone pole 10 ft away. It can do that. Gone are the days of generalized images which then required several itterations before fine tuning to what what the artist intended.

Text in image. A benchmark hurdle overcome. Art AI could not create text within image until now. The implications are obvious what an increase in creative potential this is. In other text news, another bombshell dropped. Dall-E 3 will be able to create images and then provide a creative description of that image. We at Chronicle, are patricularly interested in this detail for our own AI.

Image Credit: OpenAI - text in image, sure let’s solve that already.

If eyebrows aren’t raised now, to think what the next generation will be capable of, the adjectives to describe that are varied, and I’m betting selection is divisive. The ever faster rate at which AI is improving speaks to a next revolution. The likes of which will have no lesser a historical significance than the industrial. The fact it will likely play out in one fifth the time might also catch the attention of a few disinterested demographics.

In any event, we’d like to make sure that table-top gaming is well served by the new potential these advancements unlock.

It is plain to us that these new tools use within DND and other table-top games is not a matter of ‘if’, but ‘when’. These kinds of games are defined by their creative freedom and imaginative expression. Some people are able to map their lives along side these games as they reward a player’s investment over several years. Providing tools that can meet that level of creative freedom is usually impossible without great amounts of preparation. Moreover the tools available for preparation do nothing currently to uniquely tailor their resources contribution to a session. That is to say, if you had a castle in mind for your game, the current supplementation is someone elses description of a castle, rather than the kind of castle you were thinking of in your story. Look for resources for your game now and you just hope the things you find fit your story, or you fit your story to the things you find. Player’s of table-top games needn’t have their imagination limited to another person’s description of what they imagined.

Not sure how my request for a constellation of dragon stars became this but I’m not totally against it.

The problem: Players want to prep for their games with supplemental materials. Those supplemental materials are most often not bespoke, finite, non contextual, image separate from description. Unless a player want to spend ages in photoshop (some do and good on them) handouts for games are difficult to create given the time it requires to make them.

The solution: Radiant will generate for any DM | GM | Player, supplemental material they’d like in the form of both an image and bespoke description according to what they request from Radiant Assistant. Would you like a castle? No more looking up some random description, describe exactly the castle you’d like and Radiant will do the rest.

We’re targeting ‘note taking’ as a task many players cite as the most frustrating for their sessions. Often a DM only has time to make a brief set of notes for a description of some person place or thing they’d like to integrate in the story. When it comes time to introduce that asset in the story, with no handout to accompany it, some member of the players party will usually take notes to remember it in the canon of the story. Solving what a pain it is to generate bespoke assets fixes this problem, as every handout lessens the necessity for note taking. When you’ve a visual and written history of the story beat, there’s no need for notes.

Another cool mistake. Reminding me a bit of ‘Animorphs’ novels from my teenage years.

We would like both DMs and players free to explore any creative notion in great depth and at a rate, until now, impossible. When I polled players on what kind of cards they might make if they could freely create one, the answers could be oddly specific. One player, for example, said that she’d like a button which could generate the contents of an NPC’s pockets. This kind of granular detail is exactly the detail table-top games is best for, and now that level of detail is possible in the elements we can create for our games.

The joy of adding some unique and flavorful addition to a game session is something I’m grateful to have been party to in my life (pun intended). Seeing a friend provide some personal asset, be that an image, a written note (sometimes handwritten by hardcores), a song, or whatever a player decides to bring to the proverbial table, immerses in such a satisfying way, as to require 1st-hand experience to ever really know how good it is. With work and dedication over the next several months we hope to add more of these kinds of experiences to every player of table-top games.

Thank you for taking the time to read our developers blog.

This post has no AI contributions. It is written (poorly) by a human.



Read More
progress update Charles Jameson progress update Charles Jameson

Improving AI COMBAT FUNDAMENTALS

Last week concluded with the advancement of our campaigns through AI generative art. New functionality in the form of bespoke generated images for your campaigns rather than pre-set images chosen from a limited collage. Our AI is now able to take the the input from your campaign creation and use those details to craft a unique image to use as the title media of your campaign.

Moving onward our next priority was to deal with the loudest feedback we’ve gotten from our audience so far: the narrative generated to assist, if not at first, soon becomes repetitious and lacks variety. Equipped with that information we spent the last week improving how we used the AI to generate a variety of narratives.

We approached this problem from several directions. First and foremost we did a full reconstruct of the basic langauge we use to describe what we want the AI to do. Brilliant though it may be, greater clarity always yields best results. A return to edit and clarify in this case saw a significant improvement right away. Prior to our re-working the following results were typical.

Below are 3 exerpts taken from an earlier rendition of our combat narrative AI.

5 In the somber gloom, Connor, a shadow amongst shadows, moved with lethal intent. His weapon, a silent harbinger of fire, sought its mark. The goblin, unaware of the impending doom, continued its grotesque mirth. With a swift, merciless motion, Connor unleashed his attack. The weapon connected, searing the goblin with a fiery torment. Yet, despite the pain, the creature survived, its grotesque laughter replaced by a guttural scream echoing in the darkness.

6 In the echoing gloom, Connor, with his unwavering resolve, advanced towards the Goblin. A weapon, wreathed in flickering fire, held firmly in his grip. With a swift, practiced motion, he lunged, the fiery weapon searing through the dank air to meet its target. The Goblin recoiled, a pained shriek piercing the darkness as the fire consumed its flesh, but death did not claim it yet.

7 In the eerie gloom, Connor advanced, a weapon of cruel intent gripped tightly in his grasp. The goblin, unaware of the impending peril, continued its menacing dance. With a swift motion, Connor unleashed his attack, a surge of fire erupting from his weapon. The goblin shrieked as the flames engulfed it, the acrid smell of scorched flesh permeating the air. Yet, despite the severe burn, the goblin remained standing, its eyes ablaze with a newfound hatred for its attacker.

While in isolation a single narration can read well, when put side by side you quickly notice areas that could improve. Combat in a DND session can take hours and if you had a narrative which repeated itself, you’d rather it wasn’t present. For that reason, we made this our highest priority to fix.

Other than the basic re-write our CTO, Connor, focused on de-loading the cognitive task we’d given our AI. He provided, in the form of code, logic that the AI could utilize to better randomize its results. Remarkable to see what doing a bit of the randomizing in code could do for the overall performance of the AI. Where before it seemed to listen only occasionally to us. Afterwards, its performance was a clear line in the sand, the narrative results were better following the update. Greater variety in nearly all elements of the narration were observed following the change.

What follows are 3 exerpts following our changes to the AI

48 Stance shifted, Connor launched his fire-infused weapon at Goblin. The dark mood was palpable, the world seemingly holding its breath as the weapon, known for its scorching heat, blinding brightness, and terrifying speed, hurtled through the air.


49 Goblin, with its coarse skin, bulging eyes, and sharp teeth, was in the midst of the dark, oppressive atmosphere. Connor, brandishing his fire damage weapon, made his move. The attack hit, the goblin's features illuminated in the fiery glow as it recoiled.


50 Breath held, Connor launched his attack on the Goblin. His weapon, searing with fire damage, connected with a resounding impact. The Goblin recoiled as the fiery damage took hold, its form shuddering from the force of Connor's assault.

Describing the start, the combatant, opponent, weapon, all details of the confrontation have been tweaked to yield different descriptions and combinations of narrative.

The path to this better version had some interesting speedbumps. For a time the AI would create an end to every narrative it wrote. Everything had a concluding statement and it became stale extremely fast. We instructed it to have no comment and to stop having the need to write anything at the end. It listened to us, perhaps too well.

I laughed out loud when I read this. It wants to do what you tell it so bad, sometimes it confuses itself. What could be more human?

We enter the weekend feeling good about how far we’ve come. The newest version of Radiant Assistant is live and all the discussed changes within this week’s weekly update can now be experienced first hand be making a free account at www.radiant-assistant.com -

Perhaps the largest update since we’ve started working on Radiant Assistant is next. It will take more time than the previous to implement, but it’s no exaggeration to say the RA team couldn’t be more excited to talk about the details as soon as we’ve progressed just a little further.

Big thanks to those who support us on Patreon and for you, taking the time to read this.

Dave, Connor, Aya

Read More
progress update Charles Jameson progress update Charles Jameson

Chronicle sets up base of operations in Thailand. Begins second phase of work on Radiant Assistant.

The update is complete. Radiant Assistant moves forward with newly integrated generative art ai and gpt model 4.0

The view from Park 19 Residence. Our hotel for the time we’d search for a base of operations | Sky Towers, Jungle, Poverty

The Radiant team landed in Thailand on the 14th of August. Through a Thai contact made prior we arranged to meet with a real-estate agent. She was Kind and helpful, a sign of the so far consistent quality of character observed in the way Thai people treat others. I’d chosen a location for our hotel based on proximity to a transit line, which I assumed would make the house hunt easier. My first culture shock was when I discovered that particular transit line could only be accessed from specific entry points. Entry points a significant distance from where we were staying. The alternative, BTS skytrain line, was about a 25 minute (30 minute depending on who you speak to) walk in the other direction. Ekkamai, as it’s called, became our home station for the time spent searching for accomodation. And search we did, along with other life essential activities like getting bank accounts setup.

The walk to Ekkamai station. Always dotted in the Sukhumvit area with massive condo buildings.

To not erode the pace of the post I’ll only briefly pause to mention the noteworthy Thai banking experience. With a masterfully simple implementation of qr code banking made available to the population, it’s safe to say in my experience of both Japan and Canada, both are losing ground to a progress hungry nation which has managed to leap-frog their complacent banking systems. If Thailand can do this…

Getting a chance to walk around near our first hotel the colors of bangkok were like none in a city I’d seen before.

Returning to the hunt for our home, a few sleepless nights yielded an excellent selection of apartments. Searching through the several real-estate sites on-line gave me an insight and means to assist our agent who was making the calls and booking us the viewings. We narrowed down where to live based on proximity to at major BTS station, quality of building and its surrounding neighborhood. We spent a little extra time searching for just the right one, but are confident for having done so that we made the right choice. Where often the buildings we visited had good value in the rooms we saw, many were noticeably aged. The one we found was built in 2018. Where buildings advertising pools and gyms sparked excitement at first, many once viewed, failed to live up to their hype. That is again, until we found our building. It’s gym was actually good. It’s pool is surrounded by a well maintained garden and the infinity design allows for views of bangkok sunset. You’re free to swim in the evening and watch as the city comes on with lights. It’s one of my favourite first memories since arriving here. Of course the true center of focus has to be the affordability. Living in a 40 sq meter condo built in 2018 with the aforementioned amenities most places in the world these days will cost you around $3000, probably more actually. Vancouver, where I spent my mid-twenties, has become a world-wide meme for hopeless cost of living. The world is united in its scarcity of housing these days and to find a place where your dollar goes a little further, is a safe harbor indeed.

Through the mud up to heaven the lotus finds a way.

With our housing taken care of, finding a rhythm and refocusing on Radiant Assistant became the priority. Working in a new setting, just having moved, all the new sights and sounds, smells and flavors, added up to a blur of a first work week. Speaking together and dedicated brainstorming yielded the usual actionable results. We agreed on where we should focus.

Ambitiously we chose to see if we could get generative art working in our campaigns. A task no one on the team had any experience with and likely would involve some need for aws cloud hosting for storage of assets we generated. You know, something easy.

A huge shout out goes to our CTO, Connor, for going in blind, working hard for the last two weeks and giving us real results to now share with the community. Where before when you generated a campaign, depending on the biome you chose, a pre-made image would be assigned to that campaign when it was made. Now, given the hard work of the Radiant Assistant team you can expect each campaign created to have its own unique image. Each tailored to the specifications you created it with: biome, weather, and mood.

Never see the same campaign twice. Each image is custom generated according to your choices.

While you may occasionally create a campaign from similiar choices, you’ll never be faced with the same experience. If you choose forest for your biome, stormy for your weather, and epic for your mood, each time will yield a satisfyingly different result. It was a lot of fun playing around in our testing to see what kind of imagery the different choices prompted. Subtle changes in tone for the campaign or more obvious like weather were plainly felt in each generated piece of art.

Our campaigns, however, were not the only part of Radiant Assistant that saw a major improvement over the past two weeks. Until recently Radiant Assistant used the open ai gpt 3.5 turbo LLM AI to power its narrative capablilites. We are excited to announce we have now upgraded to gpt model 4.0, a 10X increase in intelligence. Said more plainly, when Radiant assists you in telling a story, it’s become far better at it. Where the previous model could impressively create narrative for combat scenarios, it did struggle with nuance and connecting larger more complex ideas. The vast improvements in the gpt 4.0 model allow it a far greater command over any kind of narrative task it is given. For example if you were to request a full story from the gpt 4.0 model it would do so with minimal effort, where the gpt 3.5 turbo, would have struggled to maintain coherency.

Franz has 100 million downloads attributed to the projects he’s designed on.

Where we’ve made strides in our technical developement the community surrounding Radiant Assistant continues to grow, connect, and offer invaluable assistance. The addition of which creates the potential to exceed what are limited team size is capable of. Franz responded to our post asking people to contact us if they had an interest offering feedback on Radiant Assistant’s development. His insights into design have already caused a directional shift in how RA will look and feel going forward. Franz will also feature as a guest on my podcast ‘The Interstice Podcast’. You can hear more about his career path there if you’re interested in design.

The vibrance of the nieghborhoods is striking.

User feedback is Crucial as well and we’ve had some excellent data from that. Not a single system in the website hasn’t been updated due to user feedback and our approach going forward will remain the same. We plan though, to improve how we are collecting that feedback. Our current system invites you to write back at length to us in the form of a ‘homework-assignment-like-email’. We’ve made plans to re-work this point of connection with the new users in the community and we will make a google form to accomplish that same effort. But now you can focus on what you want to say, rather than spending 30 minutes typing an email, when you have other things to do.

Slaving over the hot desktop in downtown Bangkok

This last month was a blender of change. You feel a bit dislocated after uprooting and going somewhere new. That said, with the dislocation comes a sense of satisfaction as well. Nothing to make you feel like anything is possible than setting up life somewhere new. I am glad to have this opportunity and to be able to share that with the team here. Everyone has worked hard. I look forward to sharing more of the team’s highlights and achievements as Radiant Assistant cracks on with its development. Thank you for reading and never hesitate to reach out. Collaboration is the lifeblood of Chronicle and we look forward to the next great connection.

Read More
open ai, progress update Charles Jameson open ai, progress update Charles Jameson

Radiant, The world’s first story-telling ai assistant

As humans it is in our nature to identify patterns. It’s that ability, which to some degree, allows us the occasional accurate glimpse of the future. More predictably, if your idea of seeing the future is knowing you’ll begrudingly pay ten dollars for a coffee in the following week. It gets a bit more hazy when considering where technology may take us in the next several years. It more usually appears to us as chaos. But, and it’s a big but, sometimes, we catch a glimpse past the veil and our observation of something greater coalescing is accurate.

How the team at Chronicle came together is one such event that has me suspiciously looking over my shoulder checking for a smirking old figure with a white shaggy beard and sandals. Maybe he’s walking in a sunbeam. Too many random moments of serrendipty have occured, but then again, I am a pattern seeking primate and it could all be smoke and mirrors. Regardless, when an opportunity presents itself that aligns with every goal you’ve ever had in life, you either take it, or look back forever with the regret of knowing, on whatever terms, you were offered your dream and rejected it.

I, 'Davetrippin' by my Youtube alias and Charles David Jameson by another name, had until recently lived in Kyoto for the past year and a half... attempting to work on a borrowed dream. It would not come to pass and life compelled me again to move and make a new plan. I had always thought if not for the cinema direction that coding was something could hold my interest. To give some background my dad was a 1st generation programmer and wowed me as a child with those skills in the form of hacking the first PC games that were ever released so that I, a child utterly obsessed with games and reading, would have a never-ending supply of new experiences to enjoy. As time went on though I found my inability in key topics such as math would prevent, or at the very least, form a daunting barrier between me and ever pursuing the art of coding. I focused instead on pure art. But now at the end of Kyoto and a decade dedicated to filming, what should I do?

I should attempt to code, my deficiencies be damned. I would do this by registering for a popular coding bootcamp in Tokyo by the name of Le-wagon. I studied for several months beforehand and felt like I’d be in a strong enough position to not be left in the dust once it got started. Oh boy! Was I ever wrong. My lack of technical intelligence once again reared its ugly head and I was again forced to confront the fact that when things get technical, my head, only barely metaphorically, explodes. There are times in life when it gets hard, and you have a choice, you either give up, or keep on going. This scenario and I are good friends. Sometimes it feels like an abusive relationship but what I’ve learned from the wounding is how essential it is to developing the character required to make a dream into reality. This is a long winded way of saying I’m a few sandwiches short of a picnic when it comes to the new skill-set but I’ll give it my best anyway.

Time passed at the bootcamp and one day, while in line for the bathroom, I heard a voice behind me, “Hey, you’re Davetrippin?”. I wish I could tell you I turned and said stoically in response, “I was”. But instead opted for the much more elequent, “ya?”. With that upward intonation like my answer was a question. I was addressed by the one and only, Connor Alexander Minto. He had, much to my surprise, seen some of my old videos. We fast developed a rapport. Connor would lead the class in all things coding and I would accost him for help on the various activities we worked through.

Where Connor dominated the coding, my breakthough came in the form of an unexpected turn in the curriculum. I had read it in full before beginning the camp but somehow missed the part where each must present a pitch for an app to develop. Everyone in the class (32 students) would pitch, and of those pitches, 8 selected to build. Those 8 determined by each student voting on the projects they wanted to work on. I may not be the best coder, but if there’s one thing I have more than enough of, it’s ideas, so I got to work on a pitch.

I had used chatgpt a lot before coming to the bootcamp as a mentor to help me learn coding. During that time I observed the potential that this new technology had for assisting story-tellers and writers. But exactly how, I was not yet sure. I needed another brain. Immediately I knew who to speak with: my extremely skilled dungeon master friend. Perhaps he could think of some pain a narrative ai assistant would heal. I was not wrong. He identifed the struggle of telling a good story in dungeons and dragons while in combat.

I’ll pause here for the uninitiated and give a brief explanation of dungeons and dragons for context. The game is essentially a shared story-telling experience wherein one God nerd, the dungeon master, leads the story and provides the rules for the world. Time is split between a few different activities that move the story forward. Exploration is the most straightforward. During this activity players make their way through the world and the soul focus of the dungeon master, hereafter referred to as the DM, is to tell a good story. A decent narrative is easiest to establish here as there’s little else to consider. That all changes when you enter combat. There are a million rules to consider, who’s turn it is, dice rolls and much more. Telling a good story during that time can become difficult for the aforementioned reasons. My friend observed that if something could assist him telling the story in that critical time it would be tremednously helpful. Radiant was born in that moment.

I would aim to create an ai assistant which would assist DMs in their effort to tell a good story. I worked hard on the pitch putting in extra hours honing it, practicing it. We had 3 minutes to present and if you hoped to win hearts and minds to your cause it needed to be watertight. No room for error. Pitch day would come all to quickly. I was excited until being informed that PowerPoint, the software in which I created my pitch, was outdated. Everyone was going to use google slides. Being assured that the computer we must use for the presentation might not work with the PowerPoint slides I’d prepared spiked my cortisol levels. What could I do other than keep going?

Shot through with adrenaline I began my pitch. Time flew by and the intial lead-in felt decent. I reached crucial point whereon the next slide I would introduce the app and it broke. No one even saw the name of the app. 1 minute had passed. We then spent the next two minutes fixing the program and managed to succeed just as my time ran out. I tell you candidly now, I was crushed. All that effort and excitment to be flattened by the brick wall of circumstance. The presentations continued as did my slide into a defeated posture in my chair. Eventually people began to vote and we finally reached the moment of truth where all voted for projects would be revealed and the 8 selected shown on screen with the corresponding team that would work on each.

Second to receive enough votes was Radiant.

What! I hear you say, dear reader. How could this have happened? Connor provided me with the answer afterwards: “Dave, you didn’t need to do a pitch. You wouldn’t shut up about it for the week before we presented. You kept asking people for feedback on what could make it better. You didn’t need to say a word. We all knew what you wanted to make,” he said. Humbled, I must admit it was difficult to enjoy the moment. I was so wrapped up in my woe that such a sudden change in the course of events took me totally by surprise. I recovered though and made every effort to make sure the members of my team knew they had my complete dedication and appreciation for having chosen my pitch as the one they’d like to build.

What followed was two focused weeks of bringing this pitch from concept to reality prototype to presentation day. I am tremendously proud of what the team accomplished in that time. I knew from the moment we started working, as much as any man can, that this was something I wanted to take further.

Life makes sense when we look back, and when I look back, to ignore the confluence of events that led to Radiant being built would be, in a word, stupid. A man I respect a great deal for his remarkable presience and belief in the strength and beauty of the human spirit is Ray Kurzweil. Currently head engineer at Google he is perhaps better known for his books, one specifically, ‘The Singularity Is Near’. Among the many points I’d cherry pick from what Kurzweil has to say on the topic of inventing something, the one which spoke most to me identified how invention is a matter of two things: idea, and timing. You can have the best idea in the world but if the timing is wrong there’s not much you can do about it. Nicola Tesla might have a few things to say on the topic. Radiant feels like this kind of crossroads. To be the first at anything in the modern context seems a fool’s game and mostly you’d be right. But always history grants us new opportunties, in this case, in the form of technological progress.

Large Language Model AI marks a moment in history for those who can see the opportunity, to do something that has never been done before. We were the first bootcamp to attend Le-wagon with access the LLM api. I had always in my life been interested in coding but never able to do so until this moment. That combined with meeting Connor, to me, is a kind of magic. And I can’t wait to see what kind of spell we can cast.


Thank you for taking the time to read this post. Through this channel we will communicate with the community weekly. We will touch on all aspects of development and as well share our journey as a team. If my story piques your interest and you too are a seeker of glorious pioneering innovation, then please do consider filling out our alpha tester application. Radiant is nothing without the community of people who have so far contributed to its progress and we firmly believe this should continue to remain the case. Radiant is for everyone, the spirit of it, and the goal. We are all of us characters in this great story called life, and a story is meant to be shared.



Read More