Anthony Panozzo
LightNews LightNews
panozzaj.bsky.social
Anthony Panozzo
@panozzaj.bsky.social
240 followers 460 following 530 posts
Here for AI and more positive vibes
Posts Replies Media Videos
panozzaj.bsky.social
Interesting post: The Sugar Loaf Test: How an 18th-Century Ledger Reveals Gemini 3.0’s Emergent Reasoning open.substack.com/pub/generati...
The Sugar Loaf Test: How an 18th-Century Ledger Reveals Gemini 3.0’s Emergent Reasoning
A deep dive into my experience testing the new Gemini 3.0 Pro and the growing evidence I’ve seen for emergent neuro-symbolic reasoning.
open.substack.com
November 19, 2025 at 4:19 PM
Reposted by Anthony Panozzo
cpaxton.bsky.social
A good day for semi-humanoids: itcanthink.substack.com/p/a-good-day...
A Good Day for Wheeled Semi-Humanoids
New Robots from Tangible and Sunday Robotics
itcanthink.substack.com
November 18, 2025 at 3:33 AM
panozzaj.bsky.social
Unclear how the percentages would not be 50%?
November 18, 2025 at 11:34 PM
Reposted by Anthony Panozzo
emollick.bsky.social
I had access to Gemini 3. It is a very good, very fast model. It also demonstrates the change from chatbot to agent. www.oneusefulthing.org/p/three-year...
Three Years from GPT-3 to Gemini 3
From chatbots to agents
www.oneusefulthing.org
November 18, 2025 at 6:57 PM
Reposted by Anthony Panozzo
simonwillison.net
My notes on Gemini 3, including analyzing a 3.5 hour council meeting audio recording and performance on a new, improved version of my pelican on a bicycle benchmark simonwillison.net/2025/Nov/18/...
Trying out Gemini 3 Pro with audio transcription and a new pelican benchmark
Google released Gemini 3 Pro today. Here’s the announcement from Sundar Pichai, Demis Hassabis, and Koray Kavukcuoglu, their developer blog announcement from Logan Kilpatrick, the Gemini 3 Pro Model C...
simonwillison.net
November 18, 2025 at 7:09 PM
Reposted by Anthony Panozzo
timkellogg.me
Gemini 3 DeepThink is 2x SOTA on ARC-AGI-2
A dark-themed scatterplot titled “ARC-AGI-2 LEADERBOARD.” Axes • Y-axis: Score (%), ranging from 0% up to 55%. • X-axis: Cost per task ($), logarithmic from $1e–3 to $1K. Legend Top-left: ARC Prize | Verified Data Points Colorful triangular markers represent many different model variants. The plot shows a strong cost–performance frontier: most points cluster below 10% score, but a few models stand far above. ⸻ High-scoring outliers (upper-right and upper-mid): • Gemini 3 Deep Think (Preview) — ~50% score, very high cost (≈$500–$1000). Green triangle on the far right and highest on the chart. • Gemini 3 Pro — ~30% score, cost around $1. Green triangle centered horizontally. • GPT-5 Pro — ~17% score, cost ≈**$10**. Blue triangle above the mid-cost range. • Grok 4 (Thinking) — ~14–15% score, cost around $1–$2. Pink triangle slightly below GPT-5 Pro. • Claude Sonnet 4.5 (Thinking 32K) — ~13% score, cost ≈**$1**. Red triangle near Grok 4. • GPT-5 (High) — ~10–12% score, cost around $1. Blue triangle. • o3 (High) — around 8–9%, cost ≈**$0.50**. Blue triangle just below GPT-5 (High). ⸻ Mid-range cluster: 1–7% score Dozens of small colored triangles (various models and reasoning modes) cluster in the low-cost region (0.01–1 dollars). This includes: • o3-mini (High) • o3 (Medium) • GPT-4.5 • o3-mini (Medium) • GPT-5 Mini (various modes) • Tiny/low-cost models from multiple families These fill the bottom-left quadrant. ⸻ Very low-cost, very low-score models At the far-left (cost ~$1e–3 to 1e–2), many points sit at 0–2%, representing extremely cheap but weak models. ⸻ Overall picture • Gemini models dominate the top (Gemini 3 Deep Think → ~50%, Gemini 3 Pro → ~30%). • No other model exceeds ~17%. • Cost rises steeply with performance—especially for the top Gemini variants.
November 18, 2025 at 8:59 PM
Reposted by Anthony Panozzo
wilreynolds.bsky.social
Bad "AI" content is just bad content, written faster.

While agents sound like wizardry they aren't, here's 6 basic Claude projects each w/ a specific function & instructions to turn ideas into quality posts.

❤️ The "devil's advocate" agent!

www.seerinteractive.com/insights/ai-...
outline of 6 agents to create content
November 17, 2025 at 2:03 PM
Reposted by Anthony Panozzo
cpaxton.bsky.social
A short post about a really cool video that appeared on the internet a couple days ago, from MindOn tech. It’s the most fluent and fluid home robot video I’ve seen, I’m pretty sure it’s real, and here is how/why.

open.substack.com/pub/itcanthi...
The New Cool Home Robot Demo from MindOn
An impressively fast, fluent, and natural home robot demo from a new company has captured attention all over the world
open.substack.com
November 16, 2025 at 2:45 AM
Reposted by Anthony Panozzo
wired.com
Super-cheap EVs exist in other parts of the world, as advances in battery tech and manufacturing are making new cars significantly cheaper. But the US market presents unique challenges for automakers. www.wired.com/story/when-w...
When Will the US Finally Get $15K EVs?
Super-cheap EVs exist in other parts of the world, as advances in battery tech and manufacturing are making new cars significantly cheaper. But the US market presents unique challenges for automakers.
www.wired.com
November 15, 2025 at 10:04 AM
Reposted by Anthony Panozzo
sungkim.bsky.social
@alphaxiv.org released quickarXiv

Swap arxiv → quickarxiv on any paper URL to get an instant blog with figures, insights, and explanations.
November 14, 2025 at 11:52 PM
Reposted by Anthony Panozzo
anamariecox.bsky.social
From this minister's op-ed: "People wanting to experience God are more likely to meet the Divine trying to shut down the Broadview, Illinois, ICE facility than we ever were in church...God does not live in our houses of worship, but...with those who are bearing the brunt of cruelty in this moment."
jackjenkins.me Jack Jenkins @jackjenkins.me · 4d
Oh wow @reuters.com got the photo here:
November 14, 2025 at 9:38 PM
Reposted by Anthony Panozzo
gergely.pragmaticengineer.com
From a CTO at a startup:

"We interview devs by giving them a task to build an app on the spot, from scratch (2x BE endpoints, some frontend.) They can use AI, ofc - and we dig into why they did this or that.

What is surprising: 14/15 devs from Meta failed this screening."
November 14, 2025 at 10:53 PM
Reposted by Anthony Panozzo
timkellogg.me
Gemini 3 on par with experts

this riveting (surprise!) tale from a determined 18th-19th century historian explains

1. OCR can be difficult, excruciatingly difficult

2. Gemini 3 seems to do true reasoning to decipher old hand written ledgers, beyond GPT-5-Pro

open.substack.com/pub/generati...
If this behaviour proves reliable and replicable, it points to something profound that the labs are also starting to admit: that true reasoning may not require explicit rules or symbolic scaffolding to arise, but can instead emerge from scale, multimodality, and exposure to enough structured complexity. In that case, the sugar-loaf entry is more than a remarkable transcription, it is a small but clear (and I think unambiguous) sign that the line between pattern recognition and genuine understanding is beginning to blur.
November 13, 2025 at 1:52 AM
Reposted by Anthony Panozzo
timkellogg.me
Looped LLMs

Unlike RNNs, HRM, etc., this starts with an already pretrained LLM, adds a loop, and trains a little longer

it’s a very frugal approach to creating deeper models

github.com/mcleish7/ret...
A schematic diagram comparing a layered model architecture (left) with a modified recurrent design (right). On the left, a vertical stack of labeled blocks represents model layers from L0 to L21: • The top layers L0–L1 are in light blue, with a dotted arrow leading right. • Middle layers L2–L15 are shaded gray and marked “Removed.” • Layers L16–L19 are dark green, connected by a green dotted arrow to the right. • The bottom layers L20–L21 are red, connected by a red dotted arrow. On the right, a corresponding vertical flow shows the new structure: • At the top is a Prelude block (light blue). • Below it, a small dark green box labeled s₀ ~ N(0, σ²)ᵛᵃʳᶦᵃᵇˡᵉ appears beside e and sᵢ. • Next comes an Adapter (orange). • A large Recurrent Block (dark green) includes a looping green arrow that feeds back into itself, showing iteration producing sᵢ₊₁. • At the bottom are Coda (red) and p. The diagram overall illustrates converting a deep stack of transformer layers into a compact recurrent architecture, where early layers form the prelude, middle ones are replaced by a recurrent block, and the final layers form the coda.
November 12, 2025 at 11:51 AM
panozzaj.bsky.social
Interesting post: Building Community Tech open.substack.com/pub/jeanhsu/...
Building Community Tech
Three recent community tech builds, lessons from vibe coding, and rediscovering the joy of building
open.substack.com
November 10, 2025 at 8:58 PM
Reposted by Anthony Panozzo
danabra.mov
i'm looking for a new job
Hire Me in Japan — overreacted
I'm looking for a new job.
overreacted.io
November 10, 2025 at 3:31 PM
panozzaj.bsky.social
Interesting post: A list of other catastrophes that are probably fake open.substack.com/pub/andymasl...
A list of other catastrophes that are probably fake
Adding to this over time
open.substack.com
November 10, 2025 at 4:19 PM
panozzaj.bsky.social
The Present Tragedy Strikes Whatever happened to dhh? I regret to inform you that, after a long struggle, he eventually lost his fight against the parasitic fungus that was taking over his brain,14 and died. The exact timeline is unclear, but we suspect he succumbed to his illness sometime in 2020. A morbid side-effect of his condition is that his body is still alive, walking around, delivering keynote addresses at confs and screeds on social media. Yet he exists only as a macabre meat puppet, his sinews pulled taught, controlled entirely by the brain parasite he contracted from a WhatsApp group chat. This state of affairs is upsetting, and confusing, but it bears repeating: the man we knew and loved as dhh is gone. All that remains today is an unsettling creature better known to us as David Hamburger Helper.
November 10, 2025 at 2:03 PM
panozzaj.bsky.social
Interesting post: How we built a custom vision LLM to improve document processing at Grab engineering.grab.com/custom-visio...
How we built a custom vision LLM to improve document processing at Grab
e-KYC faces challenges with unstandardized document formats and local SEA languages. Existing LLMs lack sufficient SEA language support. We trained a Vision LLM from scratch, modifying open-source…
engineering.grab.com
November 9, 2025 at 8:58 PM
Reposted by Anthony Panozzo
asmartbear.com
Just published a new article! 👇🏾

Mission, vision, purpose, BHAG. Are they useful north stars, or academic nonsense?

Many thanks for spreading the love with ❤️ and 🔁!
Mission, Vision, poTAYto, poTAHto
Mission, vision, purpose, BHAG. Are they useful north stars, or academic nonsense?
longform.asmartbear.com
November 9, 2025 at 3:39 PM
Reposted by Anthony Panozzo
katemessner.com
Loving today's news that the mysterious "fedora man" outside the Louvre heist was actually a 15-year-old museum visitor who dresses like a 1940s French detective all the time, just because. apnews.com/article/louv...
Fedora man unmasked: Meet the teen behind the Louvre mystery photo
Fifteen-year-old Pedro Elias Garzon Delvaux has become an internet sensation after an Associated Press photo captured him outside the Louvre on the day of a crown jewels heist.
apnews.com
November 9, 2025 at 12:16 PM
Reposted by Anthony Panozzo
buenzli.dev
I wrote a blog post about Flirt, a new code review tool I started working on:
blog.buenzli.dev/announcing-d...
Announcing Development on Flirt
blog.buenzli.dev
November 5, 2025 at 7:01 PM
panozzaj.bsky.social
Watched this documentary on the Gilded Age and it was pretty good!
youtu.be/yjpYzFtxfjU?...
The Gilded Age | Full Documentary | AMERICAN EXPERIENCE | PBS
YouTube video by American Experience | PBS
youtu.be
November 8, 2025 at 6:08 PM
Reposted by Anthony Panozzo
nerdy.dev
Rad dithering breakdown / interactive learning experience

visualrambling.space/dithering-pa...
Dithering - Part 1
Understanding how dithering works, visually.
visualrambling.space
November 8, 2025 at 7:41 AM
Reposted by Anthony Panozzo
jakearchibald.com
I dug into progressive image rendering and found a bunch of common assumptions aren't quite true.

➡️ JPEG & WebP progressive-renders worse in Safari
➡️ AVIF _does_ support progressive
➡️ But JPEG XL in Safari doesn't
➡️ JPEG XL decodes much slower than AVIF
⬇️ and more

jakearchibald.com/2025/present...
The present and potential future of progressive image rendering
Exploring progressive image rendering across JPEG, PNG, WebP, AVIF, and JPEG XL.
jakearchibald.com
October 16, 2025 at 10:16 AM