Diogo B. G. Sequeira

Direito Católica Porto- #linux #monero #nickland … Or Butlerian Jihad?

[Diogo B. G. Sequeira @dbgsequeira]

transcend through intensity

[Jim Dennis @answrguy]

@lauriewired To be fair, the various "knowledge management systems" using markdown do enable concise and flexible interlinking of pages and references to sections therein, and the tooling does facilitate following those and indexing them into various forms like concept maps and word clouds.

[LaurieWired @lauriewired]

@kyokyotosh pretty sure some of my best work is in an unsaved notepad++ document called new57

[Dmitrii Kovanikov @ChShersh]

@lauriewired Reminded me of this legend

https://t.co/iiJK8s3Dsb

[Diogo B. G. Sequeira @dbgsequeira]

@lauriewired @kyokyotosh Sublime is pretty good too

[LaurieWired @lauriewired]

young man builds “second brain”

wise man writes txt file /w current date

[DROID @droidbuilds]

it’s official, vibe coding done by legend himselff😂

[Roy @usr_bin_roygbiv]

@FleischmanMena man hours != productivity karl

[Roy @usr_bin_roygbiv]

I think about the fact someone with an engineering background with omp and 5.5 is likely 10-100x as productive as someone using claude code.

Someone using claude code is likely 10-100x as productive as someone using google/stack overflow or copilot even.

Someone using google or copilot is 10-100x as productive as someone who only does books/courses.

That person is 10-100x as productive as someone not using computers at all or maybe excel.

All of these coexist in the current economy in the US.

That's not to mention the massive differential in understanding how llms, training, hosting, infrastructure, harnesses, software work generally, or how things are currently priced, the supply chain for silicon and datacenters. Everything is incredibly mispriced at every layer of the entire stack from money printing and how quickly things are moving.

This is just completely unprecedented.

[Diogo B. G. Sequeira @dbgsequeira]

“You fuckers will be forever bewildered by lights and buttons like moths to a fucking flame”

[NO1ennn @N01ennn]

A 21-YEAR-OLD FROM CHINA RUNS 300 AI AGENTS AT ONCE. THE PART THAT MATTERS ISN'T THE SPEED, IT'S THAT NONE OF THEM CAN LIE TO HIM

he opens the dashboard and shows the swarm live, 300 Kimi K2.6 agents firing in parallel, then Opus 4.8 checking every single output against its source. this is not just a faster swarm. it is a loop that refuses to stop while anything is still wrong

he pointed it at 100 EV-market companies. first pass: 12 failed. wrong revenue, dead citations, empty fields. second pass: 3 failed. third pass: zero

this is not another agent demo. it is a system that catches its own mistakes before he reads a single row

[Diogo B. G. Sequeira @dbgsequeira]

@Pandurevich @apralky As a thought experiment, how would you do it? Where would you get the “signal”?

[blank @wepautism]

@apralky They just default to the standard deviation. Confidently average. Which is why those who rely on them for any sort of output outside of the purely mechanical simply NGMI

Even if it weren't so, not sure how wise it is to become reliant on conclusions outside of yourself

[yung macro 宏观年少传奇 @apralky]

A pattern I can’t help but notice in speaking with LLMs about general life advice is that they seem to be a lot more risk averse than I am. I think the average person probably has a more concave utility function than I do, and maybe this is a case of my utility function being an outlier and not the LLM’s. This would be in line with the existing studies on this, which are inconclusive and overall probably show a roughly human-like risk aversion.

Still, the fact that my AI is noticeably more risk averse than I am (I’ve had moments of being recommended extreme compliance/caution in instances where, intuitively, I’d only expect something like 95th+ percentile risk-averse humans to be compliant/cautious, for example advising against posting something mildly provocative) makes me want to defer to it a lot less, and I suspect that at this stage it’s given me on net negative-EV life advice cumulatively.

Overall it would be intuitive for this drift to continue, given that the incentive for a higher “risk aversion parameter” in post-training should be higher than the incentive for a higher risk aversion parameter when consuming the AI’s output – edge cases of extreme harm as a result of the AI’s advice in say 0.0…1% of consumers can cause meaningful regulatory damage to the AI lab, and it’s probably fair to say that advice-giving with the normal human utility function would clearly lead to extreme harm in at least 0.0…1% of consumers.

In the same light, I’ve noticed that AIs are meaningfully less Machiavellian in giving life advice than would be expected of an average person (if both were answering honestly), and certainly less Machiavellian than would be expected of an extremely successful person. This also often makes their life advice seem negative-EV – e.g. the model refuses to suggest one manipulates somebody for personal gain even if it might be optimal – these are more explicitly pursued in post-training than the risk parameters.

These run somewhat contrary to the common wisdom that we’ll gradually converge to a world in which an increasingly large number of people either outsource their life decisions to AIs, or fall behind in life outcomes? Could we not see a different drift in the near term – where people not making their own decisions become suboptimally risk-averse, suboptimally prosocial, etc. and the opposite gets rewarded?

[Om Patel @om_patel5]

SOMEONE VIBE CODED A VIDEO STREAM THAT IS SECRETLY 100% TEXT SO IT CANT BE BLOCKED

it plays 360p video at 30fps, but theres no actual video on the page. every frame is just colored text characters being repainted on a canvas

to the browser its not media at all, its javascript updating some text

its called asciline, and here's the trick:

> the server decodes the real video and streams it as binary packed text over websockets
> the browser paints thousands of colored block characters fast enough to look like 360p
> ad blockers and autoplay blockers cant catch it because theres no video element to catch
> it streams in kilobytes since its just strings, so it runs on trash internet

since the video is literally text, you can apply css glows to it, let people copy paste a moving frame, or feed it straight to a local llm

however, an unblockable stream is also an unblockable ad as well

[gavin leech (Non-Reasoning) @gleech]

are you kidding me

[Matt Margolis @ItsMattsLaw]

no no we don’t use AI, we’re an artisanal law practice

[Rahul @sairahul1]

Anthropic and OpenAI are both telling engineers to write loops.

Not prompts.
Not agents.
Loops.

That is not a coincidence.

When the two most important AI labs on the planet independently converge on the same pattern — that is a signal worth paying attention to.

Most engineers are still thinking in terms of single calls.

Input → model → output.

The engineers winning in 2026 think in cycles.

Output becomes input. The model evaluates its own work. The loop runs until the result is right.

This is the complete breakdown of what loops are, why they matter, and how to build them ↓

[Diogo B. G. Sequeira @dbgsequeira]

https://t.co/dGbFDDdNCr

[Miles Cranmer @MilesCranmer]

This is an insane paper and I love it

https://t.co/DP8OR5NJf2

[Diogo B. G. Sequeira @dbgsequeira]

@AdvDoHardware “Exercer o meu direito de jus postulandi” 💀💀💀

[定 @de3dsoul]

This is What Peak Performance Looked Like in 2004