Pinned Tweet
Announcing the 3DO Decomp Project. The goal is to reconstruct every 3DO title. I've built the tooling to help make it happen. See what I've done so far.
patreon.com/trapexit/posts/a…
I'm not sure where people got this but I've seen multiple folks on Reddit claim I was no longer supporting mergerfs. That is absolutely false. It is actively maintained and development happens regularly. It's now used at my firm so certainly not giving up on it now.
3
141
Anyone got a private plane and can come pick me up from JFK and bring me to AUS? I'd rather not spend 36 hours in JFK...
2
105
trapexit retweeted
Another day, another disappointment. I discovered a 15+ yr old memory corruption bug in QuakeSpasm, a popular Quake source port, with AI assistance.
I spent my free time preparing an issue and two PRs.
The maintainer dismissed them due to AI usage and decided to copy my fix.
trapexit retweeted
As of today @ladybirdbrowser is now a Rust project with some legacy C++ that we'll continue porting over 🤓
We've made incredible progress on concurrency in our engine since adopting Rust. I believe "CSS off the main thread" is a game-changer for browser responsiveness, and we're pushing forward on it!
On the safety side, it feels great to reduce our exposure to all the classic C++ traps, both temporal and spatial. One less (but very large) thing to worry about as we enter the most hostile computing environment mankind has devised: the web...
trapexit retweeted
Rclone v1.76.0 is out!
Highlights:
• New backends: Dosya, Infomaniak kDrive
• 5 new S3 providers
• Lots of security fixes
• rclone index command
• Multi-thread SFTP uploads
• GUI out of beta
Announcement: forum.rclone.org/t/rclone-re…
trapexit retweeted
I'm happy to share that the physical release of Terminus is moving along well. I finished verifying a pressed/replicated check disc on 3DO hardware, and the production job will begin next week. I'm tracking mid to end November for the release since I'm waiting to receive...
Dear @OpenAI @OpenAIDevs
I don't use Gmail. I've dismissed this like 20 times. Stop spamming.
1
100
lolz
Are you telling me that releasing uncompetitive models and then cutting usage in half wasn't popular with users?
Day 25/
We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.
We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage.
And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
3
106
And if it wasn't for me pestering him suggesting modern AI could help him do RE work on the retro hardware he has and always complained he didn't have enough time to work on... this almost completely useless driver would never have been created 😆
2
112
Don't you just love when someone you don't know DMs you out of nowhere, insults your work (in part out of pure ignorance), then asks you for help / to collaborate?
I don't hate AI but I do hate the amplification effect it has on some people's bad manners/behaviors.
2
124
This is embarrassing.
Your browser & the web server it talks to does this. Has forever. CPUs are faster than networks.
Coding is solved. The singularity is coming. The sky is falling. But neither the clankers nor the humans using them thought to use a 30+ year old technique?
We keep improving performance across Claude, and you may have noticed it starts responding sooner!
Claude Code now gzips the conversation it sends on every turn, so it reaches the model faster. Cloud sessions also start up as soon as you start typing.
1
114
trapexit retweeted
Lot's of development updates! FMVs, new bikes, band music, nitro boost, better graphics and a bunch of small fixes.
#indiegame
Presenting Opera Boot. A new boot ROM for Opera 3DO emulator to replace the need to use retail ROMs.
patreon.com/trapexit/posts/o…
ermahgerd!!! They could learn how to copy data and executables/code around like malware has done for 50+ years. Who would have thought?!?!
How are people taking these guys serious? Most obvious thing in the world. And it would be no more challenging as any other malware.
Anthropic's Frontier Red Team believes small AI models may now be months away from being capable of self replicating without access to a datacenter.
Quoting from the blog:
'The capabilities that make AI models generally powerful coding agents also generalize to cyber tasks, including autonomous replication. Like coding agents in 2024, we’re seeing signs of life for AI worms. We don’t yet have a full rogue replicator, but agents are becoming increasingly capable of completing the individual steps in the replication chain (and passing on their adaptations to future copies of themselves).
At the same time, increasing amounts of inference compute are now available both inside and outside of datacenters (see Appendix D). It’s uncertain exactly what form autonomous AI replication will take. It could be a slow, distributed “digital slime mold,” living on relatively insecure compute. It could also be a powerful cyber model that breaches datacenters at a high pace, or a model that scams people for compute credits.
As a prediction that goes beyond our experiments, we expect a future awash with agents running on all types of hardware, including some that aren’t under human control. While it’s true that IoT devices have been vulnerable for a long time, lightweight agents may be much more willing than human attackers to put in the effort to compromise large fractions of the world’s IoT compute. The rise of slow replicators would also be a bellwether for more aggressive, faster ones that can breach more hardened machines or be effective API key thieves. We should aim to harden our systems to make it difficult for such replicators to spread, but also prepare to deal with a world where they are here to stay.'
1
4
271
The Need for Speed for 3DO recomp code now released.
patreon.com/trapexit/posts/3…
Announcing the 3DO Decomp Project. The goal is to reconstruct every 3DO title. I've built the tooling to help make it happen. See what I've done so far.
patreon.com/trapexit/posts/a…
AutoBahn Tokio is now in manual testing and packaging phase. I'm backlogged on releases due to busy schedule but they are coming.
Next up: Psychic Detective (Disc 1).
Thank you to all the supporters of the project.
Announcing the 3DO Decomp Project. The goal is to reconstruct every 3DO title. I've built the tooling to help make it happen. See what I've done so far.
patreon.com/trapexit/posts/a…
3
182
I had actually done this a while ago but only now got around to upstreaming. All software emulators for the 3DO hacked the boot process. Unsure if it was to speed it up or didn't fully understand it. Either way... fixed.
github.com/libretro/opera-li…
4
182
People are making AI videos about mergerfs. Goes well with the AI generated site on a domain someone squated. 😆
youtube.com/watch?v=7oaTp_u1…
162
trapexit retweeted
Stages of grief about the effect of AI on open source.
Denial:
- AI code is slop! Nobody serious will use this.
- Open source will keep working exactly like it always has.
Anger:
- These AI slop PRs are ruining GitHub!
- Ban AI-generated code!
- If you close PRs, you're not open source anymore!
Bargaining:
- Okay, AI can write code, but only if a human reviews every line.
- Maybe adversarial AI review will make arbitrary external code intake safe again.
Depression:
- Wait, if code generation is basically free, what is a contribution anymore?
- Is GitHub-style open source culture just going away?
Acceptance:
- Open source is still open source, even if the contribution model evolves.
- Good issues, test cases, benchmarks, specifications, and ideas become the scarce inputs.
- Upstream may stop accepting arbitrary code while forking becomes much easier.
Bonus stage! Reconstruction:
- Projects redesign around maintainer-controlled agents, issue-driven development, stronger test/validation infrastructure, and new ways for contributors to earn trust.

















