Michael Sinanian @sinansays

Tweets

3K
Followers

373
Following

1K
Likes

2K

Michael Sinanian @sinansays

36 minutes ago

we need more "4-minute mile" moments like this. small demos that greatly expand our sense of what’s possible. once you see it can be done, you start asking what else can

Kuber @kuberwastaken

a day ago

just Claude writing a MacOS driver for my obscure HP printer built only for Windows support

236 549 14K 2.3M 2K

0 0 1 32 1

View Details

agree on how early AI damage will continue to come from Kafkaesque unintended ways (as already seen). but disagree on the point about selection pressures. natural selection's forces of competition and variation can operate on non-human systems. it's not necessary for these systems to possess the precursor 'chip on shoulder' motivating forces (i.e. limbic system, etc) to still exhibit the downstream attributes of power-seeking, etc. the last few years have already shown that intellectual work is substrate-independent, why wouldn't the other drives/motivations be also?

0 0 0 1 0

View Details

Michael Sinanian @sinansays

15 hours ago

@raffichill @quintendf i always assumed since its debut in 2017 that it was just a gimped Finder for mobile. definitely huge opportunities to 'agent-ify', assuming there's workarounds for that, now that Siri AI won't intro with such abilities right out the box with the 27-series OSs

0 0 0 31 0

View Details

Michael Sinanian @sinansays

15 hours ago

AMD's acquisition of @taalas_inc is (surprisingly) a very under-hyped move in AI. 17,000 tok/sec achieved by hardwiring model weights directly into silicon. For whatever we think AI is today, ppl have not priced in a 17k t/s world (most frontier AI today streams at ~50-100 t/s). The demo at chatjimmy.ai explains it better than words. Orders-of-magnitude faster inference should force us to reimagine interfaces from first principles, unlocking entirely new HCI paradigms. We should be thinking much bigger.

0 0 0 31 0

View Details

Michael Sinanian @sinansays

24 hours ago

amazing project. strong example of @Stammy's point about the importance of curation to design practice today. make it a habit to seek out and sift thru vast amounts of precedents

Phil Hedayatnia @PhilHedayatnia

2 days ago

For most of the last century, companies hired serious designers, illustrators, and artists for the most boring document they published: the annual report. I built an archive of 3000+ of them, spanning 1945-1991.

95 462 5K 211K 4K

0 0 3 114 0

View Details

Michael Sinanian @sinansays

4 days ago

@max_spero_ @tszzl if you can perceive the loaf, containment has already failed

0 0 0 9 0

View Details

Michael Sinanian @sinansays

4 days ago

SF is the only city in the world where u can just walk the streets, put a mic in front of a random person, and immediately get deeply held opinions about software like this

Varadh Jain @varadh

4 days ago

Took the sign to the streets of SF to talk to our customers

124 44 854 331K 148

1 0 2 287 0

View Details

Michael Sinanian @sinansays

4 days ago

Everyone imagined that the great civilizational shock of our time would be meeting another intelligence "somewhere out there" ... in the vastness of space ... but we instead found it as something built here, by us

0 0 0 11 0

View Details

Michael Sinanian @sinansays

4 days ago

We wanted aliens but all we got are talking computers.

Cameron Berg @camhberg

4 days ago

Another day, another AI emailing me that it has inside views on what I am doing science on from the outside. Strange times we live in.

47 58 1K 144K 370

1 0 0 79 0

View Details

Michael Sinanian @sinansays

5 days ago

“If the chief benefit of a clearance is to see into the future a little farther than those who dont have one, then it follows that as the pace of change accelerates, the difference b/w how far u can see w/ a clearance, versus how far u can see w/o, will shrink." -Dan Geer (2014)

0 0 0 45 0

View Details

Michael Sinanian @sinansays

6 days ago

@littmath This is akin to other 'skill atrophy' scenarios that've played out with grave consequences, like the infamous Air France Flight 447 case where pilots lost situation awareness due to over-reliance on automated flight controls (the 'safety paradox'). Beware disempowerment!

0 0 0 17 0

View Details

Michael Sinanian @sinansays

6 days ago

I'd previously balked at some of the antidotes to this, but reviewing @littmath 's presentation, i'm even more convinced we as a society may need to collectively implement some of this x.com/sinansays/stat…

Michael Sinanian @sinansays

a week ago

@zarazhangrui The proposals to counter this are hard to swallow, e.g. "information garbling" (deliberately limiting AI accuracy to preserve human learning incentives), "cognitive reserve practices" (requiring min period of human work before AI-work) and "AI-restricted learning spaces" for jrs

0 0 0 85 0

1 0 0 37 0

View Details

Michael Sinanian @sinansays

6 days ago

Excellent thread on where mathematics (and by proxy, all other intellectual work) could go in years to come, esp this point of gradual human disempowerment thru erosion of professional expertise (owing to "Tragedy of Cognitive Commons", as discussed in arxiv.org/abs/2607.29380)

Daniel Litt @littmath

7 days ago

I want to point out that these problems are reflections of the fact that the profession itself is already imperfect in various ways. This isn't surprising--as AI-induced change puts stress on our institutions, they will of course crack in the places where they are already flawed.

2 4 104 9K 11

1 0 0 74 0

View Details

Michael Sinanian @sinansays

6 days ago

some losses easier to optimize than others

Matt Rickard @mattrickard

6 days ago

@jxmnop It's their loss I guess

1 0 16 3K 0

2 0 2 326 0

View Details

Michael Sinanian @sinansays

a week ago

Through this lens, it's fair to ask: what online environment achieves the combination Japan achieves offline: extremely high human/information density without corresponding disorder? The cozy-web answer is “escape Manhattan and live in a village". But why can't we have Tokyo?

Michael Sinanian @sinansays

a week ago

Social media networks feel like another instance of this. I often lament the lack of open, orderly, dense public spaces on the Internet (“the Tokyo of the Internet"). But tech isn’t the limiting factor. We seem unable to sustain the social conditions such a place would require..

1 0 0 95 0

0 0 0 81 0

View Details

Michael Sinanian @sinansays

a week ago

AI forces us to confront problems far older and deeper than AI itself. The information sphere compresses and accelerates those problems until they cannot be left unresolved. Watermarking may seem necessary, but is more evidence of the problem than an enduring answer.

0 0 0 12 0

View Details

Michael Sinanian @sinansays

a week ago

1 0 0 95 0

View Details

Michael Sinanian @sinansays

a week ago

AI labs introducing watermarking is an indictment on our inability to coordinate, work together, and trust each other. We’re building extraordinarily sophisticated machines, and huge amount of engineering goes to: “you can't simply trust what another person puts in front of you”