Aidan Kelly

Honestly, it is scary seeing fascism on the march across many different western countries. There are Portuguese elections today where Ventura (the Farage of Portugal) is likely to reach the runoff showdown.

We must resist!

I've had a first stab at making a list of GenAI ppl on here but it's proving harder than anticipated! X is still the place to be for that content.

Let me know if you're interested in being added or know anyone who would be a good addition😎
bsky.app/profile/did:...

On-Policy Distillation On-policy, dense supervision is a useful tool for distillation

Gentle intro to on-policy distillation for language models from Thinking Machines. It defines a lot of key terms used in post-training which I found useful.

Seems like there is scope using this technique to post-train LMs without need for huge compute.

thinkingmachines.ai/blog/on-poli...

Nice work!

Computer scientist Yann LeCun: ‘Intelligence really is about learning’ The AI pioneer on stepping down from Meta, the limits of large language models — and the launch of his new start-up

Really interesting piece in the FT with Yann LeCun (@yann-lecun.bsky.social). Insights into the dynamics at Meta and how the winds of AI have changed recently.

Interested in learning more about the world models he has been working on.

www.ft.com/content/e3c4...

Introducing DroPE: Extending Context by Dropping Positional Embeddings

We found embeddings like RoPE aid training but bottleneck long-sequence generalization. Our solution’s simple: treat them as a temporary training scaffold, not a permanent necessity.

arxiv.org/abs/2512.12167
pub.sakana.ai/DroPE

Listening to the news agent podcast earlier with Mehdi Hasan, and talk of Trump being re-elected in 2020 rather than 2024 may have been the lesser of two evils. On the surface, it seems relatively convincing.

Computer scientist Yann LeCun: ‘Intelligence really is about learning’ The AI pioneer on stepping down from Meta, the limits of large language models — and the launch of his new start-up

Really interesting piece in the FT with Yann LeCun (@yann-lecun.bsky.social). Insights into the dynamics at Meta and how the winds of AI have changed recently.

Interested in learning more about the world models he has been working on.

www.ft.com/content/e3c4...

Watching what's happening over on grok in terms of image manipulation is really quite worrying and a clear violation of privacy. It needs to be regulated with a big stick, however I fear this is only the tip of the iceberg of what's out there on the web.

Coding Agents Are Easy, Actually Software engineering is here to stay

There's been a lot of discourse lately about how coding agents have become overwhelming, and that SWEs who don't use them yet just won't be able to catch up. I don't believe this is true, but we do need to know what problems coding agents are solving to approach them with an engineering mindset.

A Picture Is Worth Ten Thousand Tokens DeepSeek's new OCR model achieves 97% accuracy while using one-tenth the tokens by rendering text as images. The computationally "heavy" modality turns out to be the efficient one, which should make u...

I didn't see enough news about this crazy/cool paper from DeepSeek - you take text, and convert it to imagery, and you get BETTER performance. Oh and some really neat ideas about contextual memory too.

www.distributedthoughts.org/a-picture-is...

YouTube video by The Cure - Topic A Forest

Can't stop listening to this. Mixture of synth, bass and guitar alongside this genre must have felt like it was from another planet in 1980.

www.youtube.com/watch?v=qh-j...

Exploring TabPFN: A Foundation Model Built for Tabular Data

Understanding the architecture, training pipeline and implementing TabPFN in practice

Telegram AI Digest
#ai #foundationmodel #news

Trying to understand some code in Rust and it's giving me vietnam style flashbacks to C++

Merry Christmas! 🎄

Thoughts on AI progress (Dec 2025) Why I'm moderately bearish in the short term, and explosively bullish in the long term

Thoughtful piece by my current favourite podcaster @dwarkesh.bsky.social on where we're at with AI progress and reflections around intelligence itself: substack.com/home/post/p-...

Yann LeCun being let go by Meta might have some unexpected benefits 🍿

This may be very old man of me, but I feel like there is something about the bluesky algorithm/my feed that is not sucking me in enough. I know, I know, it's sort of the point but still, I need to have some pull to check out the content and it isn't quite itching that scratch.

With the AWS outage, now‘s as good a time as any to post this old strip.

Mission Radar: emerging trends and signals in affordable low-carbon heating Innovations to improve the affordability of low-carbon heating

Link to the article piece: www.nesta.org.uk/feature/miss...

🏛️ Policy trends:
Electricity price reduction is climbing the policy agenda. Recent debate has centred on long-term ways to bring down prices for households and businesses. That’s key to helping heat pumps compete with gas boilers on running costs. (7/7)

This combined heat pumps, community energy storage, EV car clubs and solar panels to cut bills and upfront costs. (6/7)

🔬 Research funding:
UK research has shifted from optimising heat pumps to green finance, mirroring the investment trend.

Example: Rossendale Borough Council received over £2 million for a place-based decarbonisation demonstrator in terraced streets. (5/7)

IKEA: Meatballs, flat-packs and now heat pumps. The company is now offering heat pumps in several European countries, including the UK since late September, through partnerships with manufacturers such as Aira. (4/7)

💡 Venture funding:
Finance for low-carbon heating is getting a green light from investors. Global investment in 2025 rose sharply for businesses offering loans and financial products to support uptake. (3/7)

In this edition, we looked at innovations to make low-carbon heating more affordable, a key barrier in our mission to decarbonise home heating in the UK. (2/7)

Mission Radar chart (a type of bubble chart) for Nesta's "A sustainable future" mission, updated Q2 2025. The radar is divided into four quadrants representing categories of innovation: "Ventures and investments" (top right), "Research funding" (bottom right), "Policy" (bottom left), and an unlabeled top-left quadrant (representing trends that are not primarily investment, funding, or policy). A dashed vertical line separates signals in the "Policy" and unlabeled quadrant from those in "Ventures and investments" and "Research funding." Blips are positioned based on time to expected impact, from the center outwards: "0-2 years," "3-5 years," and "5+ years." Blip size indicates potential impact on home decarbonisation, with larger blips signifying greater expected impact and smaller blips signifying smaller or more niche impact. Main Trends and Findings (Blips): Venture and Investment (High Impact, Short/Medium Term): Ikea heat pumps (Small, 0-2 years) Digital platform for local installers (Medium, 0-2 years) Energy-as-a-service (Small, 0-2 years) Retail investment in green finance ventures (Medium, 3-5 years) Research Funding (Medium Impact, Short/Medium Term): Turbo air-source heat pump (Medium, 0-2 years) Net zero terrace streets (Medium, 3-5 years) Community operated energy infrastructure (Small, 3-5 years) Policy (Medium/High Impact, Short Term): Increased government publications on electricity prices (Medium, 0-2 years) Increased references to electricity prices in debates (Small, 0-2 years) Top-Left Quadrant (High Impact, Medium Term): One large, dark blip is visible in the 3-5 years range. Source: Nesta's analysis of innovation trends. Companies and research projects included are those identified in Q2 2025. Blip size and positioning determined by expert judgement.

Happy to share my first piece of analysis work with the Discovery team at @nestauk.bsky.social on 𝗠𝗶𝘀𝘀𝗶𝗼𝗻 𝗥𝗮𝗱𝗮𝗿!

The Mission Radar is a briefing series on innovation signals and emerging trends in Nesta’s missions. We use AI to scan for innovation across business, academia and policy. (1/7)

YouTube video by Dwarkesh Patel Andrej Karpathy — “We’re summoning ghosts, not building animals”

What a really interesting conversation between two of my favourite people in AI in the form of @karpathy.bsky.social and @dwarkesh.bsky.social. I wish more ppl in this space took the same approach in terms of being open to debate, challenge and overall being humble.

www.youtube.com/watch?v=lXUZ...

Defeating Nondeterminism in LLM Inference Reproducibility is a bedrock of scientific progress. However, it’s remarkably difficult to get reproducible results out of large language models. For example, you might observe that asking ChatGPT the...

The author doesn't really address the wider implications of deterministic LLMs, but can imagine the reproducibility of outputs could be very powerful in terms of trust.

Link to the blog: thinkingmachines.ai/blog/defeati...

(6/6)

Used ChatGPT in study and think mode as well as prompting it to use socratic questioning to really improve my understanding of this area. (5/6)