zhyder

I had filled out the form for this. Wish Eric stuck to this instead of the Pebble revival: it'd have a bigger market.

I don't understand how the market isn't considered big enough for any phone OEM: how can it be smaller than that of foldables? Or even if it is, isn't it still big enough, and shouldn't there generally be more sizes and form factors of phones?

It's as tho the car industry decided to only make 184" long SUVs (6.2-6.7" phones) and 200" long 3-row SUVs (foldables)... no other SUVs, no sedans/hatchbacks, no sports cars (much smaller and much lower volume). And different cars are actually hard to engineer and mass-manufacture the chassis and bodies for... in contrast a phone's HW is inherently more modular and mostly just the screen and battery need to be changed for each size.

Beautiful work! I appreciate how this robot clearly does NOT try to look like any natural creature. I don't want a future where we can't easily distinguish nature from robotics. So far humanoid robots look clearly robotic too: hope that trend continues.

All current AR glasses are fundamentally designed to see the real world with existing light, as opposed to XR (mixed reality) glasses that block out all existing light from reaching your eyes and create a real-time "passthrough" video feed. So AR glasses can't really block any real world ads, can only place an annoying overlay on them. It'd work with XR glasses but no one wants to walk around in the real world with those on.

Maybe someone will invent an electrochromic layer on AR glasses that can selectively block light at individual pixels (rather than darken the whole lens, as current electrochromic layers on some AR glasses do)... that's when RealWorldAdBlock would actually be viable.

Blended price (assuming 3:1 for input:output tokens) is 3.24x of what was stated before [1], and now nearly 5x of 2.0 Flash. Makes 2.0 Flash a still competitive option for many use-cases, particularly ones that aren't coding-heavy I think. A slightly poorer performing model can net perform better through multiple prompt passes. Bummer, was hoping 2.5 Flash would be a slam dunk choice.

[1] - https://web.archive.org/web/20250616024644/https://ai.google...

What's the SOTA for generating animated vector graphics these days? The text-oriented LLMs don't seem to be able to draw aesthetic SVG paths, and the image diffusion models do bitmaps rather than vectors. I think there's a big market for a GenAI Illustrator with After Effects so hope someone cracks it.

Now all the big 3 LLM providers provide web search grounding in their APIs, but how do they compare in ranking quality of the retrieved web search results? Anyone run benchmarks here?

Clearly web search ranking is hard after decades of content spam that's been SEO optimized (and we get to look forward to increasing AI spam dominating the web in the future). The best LLM provider in the future could be the one with just the best web search ranking, just like what allowed Google to initially win in search.

Strongly agree with the author that the original prompt is much more substantive, but I think they're mistaken that "skin in the game" is a small motivating factor for human written text. It's the entire motivation: we all want to look like we've done more (get better grades at school, or get better compensation at work) while minimizing effort. We're not incentivized to produce just the substance, coz effort scales O(substance).

Yup I directly search Wikipedia often too. Hmm Hackerpedia could work, or ReddiRank.

Is there any way to access this via the API, after perhaps some oauth from the Anthropic user account?

Much better visualization of personalization options compared to other websites. Am genuinely excited. It's like Project Ara was for smartphones, except on phones there's inherently limited space to allow for personalization, and wasting a bit of space significantly trades off battery life. A car or truck has oodles of space.

Neat idea around removable (I think?) plastic panels that're the same color all the way thru. Cheap to manufacture, dent resistant and DIY vinyl wrappable.

V v cool: first time I've seen such expressiveness in TTS for laughs, coughs, yelling about a fire, etc!

What're the recommended GPU cloud providers for using such open-weights models?

One of the things that bothers me most about advertising is that we sell attention to the highest bidder. It happened before online ads too but online ad networks 'perfected' the bidding process. And I say this as someone who's made much of my wealth indirectly off of such ad networks.

It means the expensive product/service gets your attention, coz it can afford a higher bid, instead of the one that's better for you. It also sets a high floor on prices.

To some extent the bidding process was needed when there was inherent scarcity of inventory to place information (e.g. billboards or the 1 local radio station), but there's no scarcity online. Why can't we just have a web where product/service info is listed, and people can seek that info through some search engine?

How is the latency (Time To First Byte of audio, when streaming) and throughput (non-vibe characters input per second) compared to the existing 'tts-1' non-HD that's the same price? TTFB in particular is important and needs to be much better than 'tts-1'.

Text allows developers lots for flexibility to do other processing, including RAG, calling APIs yourself and multiple chained LLM invocations. The low latency of realtime API means relying fully on one invocation of their model to do everything.

I'm super happy about the revival (I backed the first Pebble Kickstarter) and the bullet list of key features you're targeting would make my dream watch, but I hope you take this as constructive criticism that you're way off right now on this one:

- Simple and beautiful design

It's ugly, and the gap with the industrial design of today's watches is wider. I suggest contracting with a good industrial design firm to redesign the case: the case material, screen and internal electronics can remain the same.

Any provider already hosting this (similar to how many providers host Whisper for STT)? Looks like doesn't support streaming tho (same with Whisper coincidentally), but great to see open models get so much better.

I backed the first Pebble on Kickstarter. I didn't like its industrial design but voted with my wallet for the _idea_ behind it. I had high hopes for subsequent iterations but IMO it never evolved into a sufficiently beautiful industrial design. From the same era, Vector was better-looking with similar screen tech, but never captured meaningful mind-share or market-share.

I'd love to have a small watch with long battery life, reflective always-on hi-dpi grayscale screen, and classic industrial design using a metal case and small bezels. For now I'm (reluctantly) making do with a Withings ScanWatch Light.

I like that https://artificialanalysis.ai/leaderboards/models describes both quality and speed (tokens/s and first chunk s). Not sure how accurate it is; anyone know? Speed and variance of it in particular seems difficult to pin down because providers obviously vary it with load to control their costs.

As a former chip designer (been 16 years, but looks like tools and our arguments about them haven't changed much), I'm both more and less optimistic than OP:

1. More because fine-tuning with enough good Verilog as data should let the LLMs do better at avoiding mediocre Verilog (existing chip companies have more of this data already though). Plus non-LLM tools will remain, so you can chain those tools to test that the LLM hasn't produced Verilog that synthesizes to a large area, etc

2. Less because when creating more chips for more markets (if that's the interpretation of YC's RFS), the limiting factor will become the cost of using a fab (mask sets cost millions), and then integrating onto a board/system the customer will actually use. A half-solution would be if FPGAs embedded in CPUs/GPUs/SiPs on our existing devices took off

I agree that seems like the next step, but how will that help push salaries down? I thought remote work would do that much more, coz many employees would move to less expensive areas in the same country and new hiring would focus on lower-income countries.

This cost savings from remote work is what I expected to push adoption of remote work more, and I'm surprised by this reversed trend.

Such an odd rebranding. Why retire a beloved brand? How does "TV Streamer" work better in describing that smart home hub is included?

This link doesn't indicate the credits are exclusive to YC, seems they're open to all AI startups: https://cloud.google.com/startup/apply

Shouldn't the right benchmark be performance per watt? It's easy enough to add more chips to do LLM training or inference in parallel.

Maybe the benchmark should be performance per $... though I suspect power consumption will eclipse the cost of purchasing the chips from NVDA or AMD (and costs of chips will vary over time and with discounts). EDIT: was wrong on eclipsing; still am looking for a more durable benchmark (performance per billion transistors?) given it's suspected NVDA's chips are over-priced due to demand outstripping supply for now, and AMD's are under- to get a foothold in this market.

I think we overstate how useful such a system is. Visual UIs with buttons are so predictable and efficient to use, so anytime you can reach for a screen and hand-operated-input for anything complex, you will.

I've stopped using home devices like the Echo (coz of privacy concerns, esp with hotword mistriggers): now use voice only when driving the car. Maybe multimodal LLMs like GPT-4o will spawn new useful use-cases, but I think they're unlikely to be for the same use-cases Alexa the product+brand is known for.

As a long-time Mac user, I'm actually eager to switch to Windows: more hardware innovation happening with more options in that ecosystem (e.g. in dual screen laptops like Asus Zenbook Duo). Windows has a few rough edges but my usage is dominated by the browser and terminal, which are similar on both Mac and Windows. Great to see these ARM-on-TSMC benefits of performance and battery life land in the Windows ecosystem.

Doesn't matter to most people what Epic does. There's a duopoly in ecosystems for the most important general-purpose computing device for most people on the planet, and we should demand that people have more free control over devices they own. On iOS, I think that's ability to sideload apps and run other web engines.

I like that the Asus one converts faster to dual monitor mode with a built-in kickstand, and the keyboard has a touchpad.

You mentioned CO2 so I hope you're integrating active ventilation through an HRV or ERV. Integrated energy-efficient heating+cooling+filtration+ventilation is the holy grail for me.

The legal code seems like ever increasing software code, except there's no incentive to refactor or delete whole swaths of it. How do we create such an incentive? Less is more here too, perhaps more so because the legal code affects everyone's ability to avoid doing anything illegal and to hold elected legislators accountable.

I think it could be used to create a videoconferencing experience akin to sitting around a table with multiple people, while still selectively facing or making eye contact with individuals.