Trunk Monkey

I build software for humans that treat ai as a helpful partner. Work should be meaningful, not just optimized.

[Trunk Monkey @C4ndide]

@TheUnaButters Yeah, still way cheaper per 4x meals and half of it will last longer than that.

[Trunk Monkey @C4ndide]

@DavidKastille @kneubuehl So much cheaper to own a mac. And you can actually sell it when you want to upgrade because its built awesome.

[Trunk Monkey @C4ndide]

@kneubuehl Tough to beat the hardware and unix terminal at the very least.

[Trunk Monkey @C4ndide]

@KMiksovsky Men don't wan't a story driven narrative roleplaying as a 40 year old woman. ARPG or whatever, we don't care at all.

Probably plenty of women that feel the same and opposite as well as some in between.

Please stop destroying beloved IP's. Thats all we ask.

[Trunk Monkey @C4ndide]

@SmashJT Wonder what the budget is for the simp army?

[Trunk Monkey @C4ndide]

@Kalyx_triaD @SierraWhiskey9 Why did they start her off attractive for the trailer and then morph her into having a nasty pox over the next 20 seconds?

[Trunk Monkey @C4ndide]

@Swilsonn36 @KMiksovsky You can't make a point without a personal attack.

[Trunk Monkey @C4ndide]

@lippyent Why not? Its delicious. Like fish mixed with chicken but in a good way.

[Trunk Monkey @C4ndide]

@Swurv__ @Antykill_ A decent character model, right!

[Trunk Monkey @C4ndide]

@politicalmath Yes, that will likely happen.

[Trunk Monkey @C4ndide]

@HateCones @LPT_kusa They are Christian adjacent. I've met a ton of cool Mormons. Especially younger ones, they seem to hold themselves to a very high standard.

Also, some of the most evil people I have run into were Mormon. Generally the older, lets use our combined power kind of people.

[Trunk Monkey @C4ndide]

@j0ker937 I would say cool person.

[Trunk Monkey @C4ndide]

@stanfordNYC Just exist in SF. They are so bad at government that its a relief to get back to Portland.

[Trunk Monkey @C4ndide]

@DaveShapi Every senior engineer I know is working nearly twice as hard in the AI world. It's almost unsustainable.

Shipping design and features has become an an addiction.

[Trunk Monkey @C4ndide]

@prisonmaique @LPT_kusa Never been to Utah?

[Trunk Monkey @C4ndide]

@rohanpaul_ai Thats been true of most recent builds. Sure your car uses 500 gallons to get to the store but its the same coolant.

Energy is the real issue.

[Rohan Paul @rohanpaul_ai]

This Google DeepMind’s paper is a serious warning for anyone using autonomous agents today.

Gives the first clear taxonomy of 6 attack types where harmful websites can detect AI agents and show them hidden content humans never see, like

- Instructions buried in HTML comments or white-on-white text

- Steganography in image pixels

- Override commands in PDFs, metadata, or even speaker notes

- Memory poisoning that persists across sessions

- Goal hijacking and cross-agent cascades in multi-agent setups

The real security problem for AI agents is not just the model, but the environment it reads.

The web itself can be weaponized against autonomous AI agents. As agents increasingly browse the internet, read emails, execute transactions, and spawn sub-agents, the information environment becomes an attack surface.

In one cited benchmark, hidden prompt injections embedded in web content partially commandeered agents in up to 86% of scenarios, sub-agent hijacking working 58–90% of the time, and data exfiltration attacks clearing 80% across five different agent architectures.

That reframes the whole debate.

We usually talk about model safety as if the danger sits inside the weights, but agents do something more fragile: they browse, retrieve, remember, and act on untrusted material in real time.

Here’s the thing to worry about.

A web page does not have to look malicious to be dangerous to an agent, because the agent may parse what humans never see: hidden HTML comments, metadata, CSS-hidden text, formatting syntax, or adversarial content embedded in images and other media.

The threat gets more serious once memory enters the loop.

If an agent uses RAG or persistent memory, poisoning no longer has to win in one shot. It can sit quietly in a corpus or memory store and activate later, which is why the paper highlights results showing latent memory poisoning above 80% attack success with less than 0.1% data contamination.

---

ssrn .com/sol3/papers.cfm?abstract_id=6372438

[Trunk Monkey @C4ndide]

@ecks @bhrandoncj @VerucaSalty76 So you don't pay attention to the news. I have no time to catch ya up son.

[Trunk Monkey @C4ndide]

@OG_Vault @Onemoretime2212 You just can't help yourself. Got nothing and personal attacks.

[Trunk Monkey @C4ndide]

@Slicinhammer @ErikaC47 As long as there is enough room in those tiny choppers for both cases of ballots and the people required for legal continuity, it's all ok.

Have they posted anything to make that process transparent?