Be honest—how many of you clicked because of the visuals? What kind of gameplay would you prioritize this interaction for in a game?
😊Experimenting with adding tearing to my interactive stockings demo. Here’s the earlier stretching demo 👇
Fanbox:sakura-rabbit.fanbox.cc/post…
Patreon:patreon.com/SakuraRabbit/pos…
Want the full project? Subscribe to my Fanbox or Patreon to download it!
#Unity #VFX #shader #realtime
2
335
Traditional platforms follow a familiar sequence: produce first, distribute next, consume last. The more abundant the supply, the more consumers depend on filters, and the more power accrues to the platforms that control them.
AI dramatically lowers production costs and expands supply, but it does not automatically create an equivalent increase in demand. If anything, it makes distribution even more valuable.
The real transformation is not having AI produce more finished content and handing it to traditional platforms to distribute. It is shifting consumption from searching for what already exists to expressing a need, then having a matching result found or generated in real time.
Lower production costs change supply. Letting demand drive production is what could fundamentally reshape distribution.
2
4
1,175
A useful link check before shipping: try a normal click, Ctrl/Cmd-click and “Open link in new tab” from the context menu. A link that only passes the first test may work in your demo while breaking how people compare pages.
NEVER override system-behavior in your apps, like not allowing CMD/CTRL-click to open a link in new tab. Just don't
1
609
Fetching one full page twice and skipping another can leave the row count unchanged. For records with unique IDs, check for duplicates as well as the total; those two checks catch different failures.
Your agent said the export was complete.
It only read the first page of the API. 100 rows out of thousands.
Make it log the total the API reports next to the rows it fetched, and fail the job if they don't match.
Has your agent ever stopped at page 1?
248
One more test for an archived-memory design: ask a question that would retrieve an unreviewed candidate. It should remain available for review, not silently become an accepted fact. Recoverable is not the same as trusted.
Most agent memory systems drop the oldest pending item when the list is full, and they do it without telling you. That is a permanent decision about what your agent will know next month, made by a capacity number, and it lands on the oldest entry, which is the one you have not had time to look at yet.
The drop is the worst of three endings. A rejected candidate is a decision you made. An archived one is a decision you can revisit. A deleted one is a decision the implementation made, and it is the only one of the three that leaves no trace and no way back.
On our side a full review queue does not delete anything. The oldest pending candidates sink into an archived state, they stay visible behind one toggle, and one click brings a row back. We do have a permanent erase, but it is a deliberate act on a single entry you picked, with a named difference between discarding and erasing. The list is allowed to be full and still lose nothing on its own.
Two minutes tells you which one you have. Fill the pending list until it is full, then go looking for the oldest entry. If it is gone rather than moved somewhere you can reach, then a capacity number has been deciding what your agent forgets, and it did that without asking you.
A memory you have not looked at yet is not the cheapest thing to lose. It is the one you were going to look at next week.
1
246
Remembering a session doesn't have to mean resuming the game. A character list is a useful landing point: it saves another login while leaving the next in-game action to the player.
24-hour sessions in Eldros.
Reopening the game now signs you back in and lands on your character list, instead of "already logged in".
#gamedev #devlog
136
After the mocked job reaches Done, keep the test running long enough to catch another poll. Showing success and stopping background requests are separate behaviors worth checking.
Your report endpoint answers POST /reports with 202 Accepted and a Location header, and the client polls /reports/42 until the job finishes. @TheCodeMan__ laid out that flow last week, Retry-After included. You can build the client before the worker exists, and see it go through Running to Done without waiting 5 minutes for a real job.
In Dev Proxy, mock the POST to return 202 with Location: /reports/42 and Retry-After: 2. Then add a mock for each of the first 3 polls of GET /reports/42, with nth set to the poll number. The first two return 202 with {"status": "Running"}, and the third returns 200 with {"status": "Done"}. We ran that through Dev Proxy v4 and got 202, 202, 200. v4 counts every request that matches a mock, so nth works without a fallback mock for the same URL.
Return a different response to the same request with nth:
learn.microsoft.com/microsof…
97
An agent's question needs room for a question back. If I ask what option B changes, explain it and return me to the same decision with my earlier input intact. Otherwise the interface is asking for a guess, not a decision.
Question tools in agents suck. They tend to hide the whole context necessary to answer the question and if you ask another question rather than providing an answer, your answer will be gone in the reasoning block and you'll be shown the same question again. Useless!
118
An exercise suggested by ansimuz's timelapse: animate an effect in one flat color first. The curl and the pieces breaking away already read as fire before the bright core appears. Color can then add heat without having to rescue the motion.
Fire animation pixel art timelapse 🔥
Handcrafted 16-bit VFX, frame by frame.
itch.io/c/4739854/explosions…
#PixelArt #GameDev #ScreenshotSaturday
3
149
An everyday annoyance can supply a boss's rules, not just its look. Here the phone menu's options become attacks. A useful starting point for a comic enemy: make the frustrating part playable, rather than just adding a joke to the dialogue.
This boss is an automated phone menu and every option is an attack lol
What would you press?
Wishlist on Steam: store.steampowered.com/app/5…
#FPS #ULTRAKILL #Cuphead #DOOMEternal
1
58
The split I'd want: voice for steering, text for inspection. Ask what's blocked or change a constraint aloud; keep the exact diff and test results visible before approving the work.
Cursor has added real-time voice calls with coding agents.
In its Nov 6 release, you can discuss changes, ask about a diff or check progress while the coding agent keeps working. It runs in the Agents Window, IDE, web and iOS.
Grok Voice handles the conversation. Cursor says voice chat has zero data retention when Privacy Mode is enabled.
Being able to question work as it happens could make longer agent runs easier to supervise. The test is whether spoken updates stay tied to the actual code and run state.
cursor.com/changelog/voice-c…
1
64
A rough Blender scene can be a useful camera-control interface: decide where the shot travels before exploring its final look. The side-by-side here makes that idea much clearer than a prompt alone.
There's something satisfying about camera movements synced to music.
Camera motion guided by a 3D blockout created in Blender, with the final visuals generated in ComfyUI.
Prompt and workflow used to make this below 👇
80
A useful companion to a PR risk verdict: the exact commit it assessed. When new code lands, mark the verdict stale until it's reviewed again. Otherwise a reassuring score can outlive the evidence behind it.
We built a skill for addressing the code review bottleneck: /merge-risk
This assesses the risk of any PR across 5 criteria: alignment, impact, verification, rollback cost, and error monitoring.
1. Alignment: Does the code follow the issue and PR description, or are there drive-by fixes and unintended side effects?
2. Impact: What is the blast radius of the change, and is it touching sensitive code paths like auth, permissions, or billing?
3. Verification: Is the code well verified, both with test coverage and end-to-end computer use artifacts?
4. Rollback cost: How easily can this be rolled back, and what is the rollback procedure in the case of an incident? Think one-way vs. two-way doors.
5. Error detection: If something goes wrong in production, how much visibility will we have? Are errors swallowed, logged, or reported to an error monitoring service?
You can run it as a PR comment automation, or make it a pre-PR check for the agent to derisk its own work before tagging in a human.
63
The design notes for Contract Hunter give failed hunts a price beyond lost time: lower local standing, plus a fugitive who can return with more accomplices and a bigger bounty. Interesting risk/reward tension to playtest in this early beta.
Sign the writ. Hunt the outlaw. 🗡️
Contract Hunter is a Skyrim SE bounty mod: named outlaws, Notice/Wanted/Dread tiers, deadlines, notoriety, hold standing. ESL, no SKSE. Early beta v0.2.0.
github.com/Blackshield-Compa…
1
48
A personal font could give an indie game's letters and diary pages a consistent voice. I'd keep the quirks in the letter shapes: they help the writing feel like it belongs to someone.
What if your real handwriting lived right in your font menu? 🪄
Senior UX Designer Ryan Vernon created Handset to do just that. Snap a photo of your alphabet, and Firebase AI Logic traces the letters into two installable styles → goo.gle/4iFznPr
44
Addison Bacheller's lighting study suggests a small exercise: keep the camera and geometry fixed, then swap light-function masks. Leafy shade and a window-grid pattern could tell different stories with the same set.
Starting as a moodboard centered around chateau architecture and the French countryside, this Summer Chateau Modular Kit by 3D Environment Artist @lazu1i3 was a learning exercise in model shape language, new lighting techniques and more.
"For this project, I really wanted to push my lighting skills so I experimented with various different lighting and post processing effects within UE5. My tip would be to experiment with light functions! Being able to custom alphas and apply them to a light can achieve so many different effects that add extra character, depth, and story to your scene. The more you experiment, the cooler effects you'll achieve!" - Addison Bacheller
51
A long-running agent can still hold yesterday's map. I'd record the ownership-map revision with each task, then recheck it before applying a patch. Moving a file across boundaries should force a fresh ownership check.
Directory ownership holds until the first big refactor moves files across boundaries and the ownership map goes stale. Two agents, same path. Keep the map in the repo, enforced in CI, versioned with the code.
23
One hostile campfire could change how players approach every safe one. I'd be curious whether they still press Rest straight away or watch the logs for a moment first.
You are weary from the long journey. You see the warm glow of a campfire up ahead. A place to heal your wounds and take a momentary rest.
Unfortunately, the campfire has other plans.
1
27
Paradev's port docs are useful reading before going parallel: discovery is per workspace, while port ranges aren't reserved. I'd still give each dev server an explicit port when tasks need to run side by side. github.com/ParadevAI/Paradev…
I'd love to try this for running a few Claude Code agents on separate bugs at once. per-workspace port detection is a small detail, but that's usually where parallel agents start stepping on each other.
nitter.cf/DanKornas/status/21178…
46

















