Sabitlenmiş Tweet
Nathaniel
20.5K posts

Nathaniel
@potrepka
building a live coding platform ... coming soon ✧ collector @mathcastles ✧ investor @zcash @miladycult ✧ algorave ✧ special people, special places ✧ 🐢🏆
United States Katılım Haziran 2015
2.9K Takip Edilen3.5K Takipçiler

@codyschneiderxx Touch is overrated, but I don’t want the affordance to go away completely.
English

Moving fast with insufficient quality guardrails is a really dumb idea because once code quality drops below a certain threshold, lifetime value of a customer goes to zero. Anthropic does not have a monopoly by any stretch of the imagination. The only rational explanation is that they’re so far behind the competition that they feel forced to move this quickly, but that’s still a losing battle.
English

Anthropic: degraded flagship website. Apparently, an annoying UX issue irritated paying customers – and no one at Anthropic noticed.
According to the pragmatic Eng. The company moves at light speed, generates 80%+ of its production code with Claude, but quality and user experience seem to be taking a back seat.
A simple observability solution could have flagged it early.
Moving fast with dedicated quality guardrails at an Anthropic scale shouldn’t be a nice-to-have.
English

we're talking about a cross-border legal battle with a company more than ten times the size. it's nice to think there are ethics in business, and there should be, but in business, you should know better than to assume that people will play fair.
they will probably settle this out of court. meanwhile, there's no point in displaying any name in the user interface until the terms are crystal clear.
English

whether this is true or not it's going to cause every company producing open source models to re-evaluate if they should continue to do so
that is incredibly frustrating
sumit@sumitdotml
now a deleted tweet, probably nothing
English

Very small set of opinions. Very small set of rules.
Very large set of possibilities.
This is “malleable software” is in a nutshell.
Generating this software with agents takes the abstraction one step further, but it doesn’t elimimate it entirely.
It’s still a necessary intermediate step, if not for the human, then for the agent.
To say that “malleable software” will be a term of the past is to make liguistic assumptions about what “malleable” and “software” mean.
You assume that “malleable” refers to buttons and sliders, but it can refer to voice or text conversation, too.
You assume that “software” refers to perfectly deterministic code, when large language models are software, too.
The term “malleable software” is, in effect, the north star of home computing. We the technical individuals seek to give non-technical individuals equivalently expressive access to computer technology as we have.
It’s a matter of translation, and we do it for ourselves, too, and in reverse for the computer hardware, when we write code.
English

@TylerDurden Any thing you use a computer for, you can tell an agent with the right integrations and context to do for you.
English

@catalinmpit Maybe not write it yourself, but Claude Max is certainly not worth the $200 anymore, if you know how to code. There's enough open source tooling for Claude to be obsolete in 6-12 months, which is why the team is shipping like crazy.
English

Lately, Claude makes some shocking mistakes.
⟶ Implements overly complex code
⟶ Ignores the codebase's code style
⟶ Removes working code for no reason
⟶ Replaces code that's out of scope from the task at hand
It feels like it needs 100% supervision. At this point, you're better off writing everything yourself.

English

@amishescapee If only there were some sort of community or religious group where divorce is explicitly prohibited…
English

@IterIntellectus severe disfigurement, severe disorders, severe iq drop, heavy mental impairment, debilitating physical defects. all of these are valid reasons. keeping such a child in the world is unfair to siblings and a drain of resources that could be used for fit children. brutal truth
English

absolutely fucking disgusting
i'll never understand what pushes someone to want to kill their own child.
but the worth of a culture can be measured by how it treats its future, and britain just voted to decriminalize murdering it
it's over for these people.
whatever suicidal ideology drove a dying civilization to this point cannot be eradicated soon enough
Dr Rahmeh Aladwan@doctor_rahmeh
The UK House of Lords has just legalised abortion up to birth. Women can now end the life of their unborn baby at any stage, for any reason, without legal consequences. A truly dark day for Britain.
English

Abortion remains legal with no gestational limit in several US states. What a truly disgraceful society we live in…
Abortion is child sacrifice performed in broad daylight.
The abortionists brag about how many abortions they perform, and the children’s body parts are sold on the black market. It’s sickening, and many young people are totally misinformed on the horrors of it all.
English

Take your favorite crypto app and slap a 2d or 3d game UI on top of it
Most dapps completely expose all their APIs on the browser so easy to reverse engineer it
zac.eth 🧙🏻♂️♦️@zacxbt
bridge swaps are working! looking for 20-30 early testers before public launch tomorrow dm, reply, or follow @bridgedgg 🌉
English

Deterministic compute is yang to the model's yin. That's why we're seeing so much excitement around harnesses.
What I'm not convinced of is that multiple well-chosen passes of a well-trained 9B model is necessarily less powerful than one pass of a 90B model, especially since the 90B model is monolithic and trained by a singular entity, while the 9B model is agile and can be replicated and fine-tuned to particular tasks.
English

thinking out loud. every model gets math wrong. 7B, 9B, 70B. doesn't matter. pattern matching is not computation.
hermes agent has code_execution which spins up a full python sandbox with RPC over unix sockets. powerful but heavy. a 9B isn't going to navigate that reliably for basic arithmetic.
what if there was a lightweight calc tool built in. model hits a math question, calls the tool, gets the exact answer computed on your hardware. no interpreter overhead. sandboxed. simple enough schema that a 9B can call it every time.
the accuracy problem stops being a model problem and becomes an infrastructure problem. and infrastructure is solvable.
@Teknium would this belong in hermes agent or is code_execution enough?
English

I believe Anthropic will beat OpenAI in the long run.
Infrastructure battles are the bigger battles.
OpenAI is betting on Python and Anthropic on Typescript.
Maybe OpenAI chose Python because they believe the web will go away.
Either way, both of them are purchasing companies in these ecosystems completely unrelated to AI.
English











