j

super cool to see @orhundev reviewing my 3D volatility surface at @ratatui_rs. never expected to see someone reviewing my pr on youtube (episode 15 of ratworks). also big thanks to @joshuamck

@ilaffey2 claude find oil refineries and power stations on google maps api

i’m the fucking jordan belford of fixed wing drone software

ICS-24 keyspace demonstration

ICS-20 token transfer dynamics

Some visuals from my Builderfest presentation 🧵 : Tendermint consensus

@Ryun1_ @Cardano_CF @txpipe_tools @diegomac35 @buidler_fest inter blockchain something idk ask @CryptoJoe101

@0xfloorlicker @Cardano_CF @txpipe_tools @diegomac35 @buidler_fest what does IBC mean

excited to finally make it to buenos aires this march. been working on something big, more details coming soon.

Builder Fest #3 | Argentina Edition A room full of talent, conviction, and builders shaping the future of the ecosystem. It is a true honor to share this space with some of the brightest minds in Cardano. Hosted by TxPipe, with Santiago Carmuega, Diego Mac, Federico Weill, and Rodrigo Suaya, together with Lucas Macchiavelli and Mauro Andreoli. Supported by Input Output Global, the Cardano Foundation, the Midnight Foundation, Intersect, and the Cardano Ambassadors Program. Fully endorsed by Charles Hoskinson and J.J. Siler Ghost chain? Then these must be the most brilliant ghosts I have ever seen. You do not get rooms like this in a ghost chain. 👻🔵

@coc_space Frederik mentioned IBC and Hackathon, which will happen at the 2026 @buidler_fest I and Julius will run a presentation + hands on session. x.com/0xfloorlicker/… For all the dev on the space attending, come say hi and check all the good things we're working on

@_dominatos polyfill-rs 0.3.1 will be coming out later this week, has gotten even faster with GTC/GTD/FOK/FAK order support, just doing some final tests.

@CryptoJoe101 @buidler_fest @Cardano 🇦🇷🇦🇷🇦🇷

📢 PSA I will be attending @buidler_fest and will be serving 🫖 and ☕️ while that legend of @0xfloorlicker will give a hands-on presentation about @Cardano and the IBC bridge.

@aiktp_com @Cardano_CF @txpipe_tools @diegomac35 @buidler_fest hope to see you there!

@chaumian only semi-related, but we had a similar approach to investigating tick-wise invariants for CLAMMs like Uniswap V3: arxiv.org/abs/2512.06203

A Formal Approach to AMM Fee Mechanisms with Lean 4 arxiv.org/abs/2602.00101

@orhundev @ratatui_rs @joshuamck np hopefully more incoming

@0xfloor_licker @ratatui_rs @joshuamck haha thanks for the PR! it's amazing

super cool to see @orhundev reviewing my 3D volatility surface at @ratatui_rs. never expected to see someone reviewing my pr on youtube (episode 15 of ratworks). also big thanks to @joshuamck

@SrinivasanSS52 @orhundev @ratatui_rs @joshuamck it was a demonstration of perspective projection in the terminal, this case for a 3d volatility surface, but the same model is also used in GIS data for example. felt like people using ratatui were missing out on this kind of use-case.

@0xfloor_licker @orhundev @ratatui_rs @joshuamck what was the PR about

@jibranhutch is it something you can benchmark with the same tools i.e "verifier-free verifiability" that isolates implicit formal reasoning from explicit RLFR shaping, V@1 no logs, +reflection-only, sweep vs compute. check %VCs vs tests/spec-mutate to avoid solver hacks. could be worth it.

@0xfloor_licker Yeah we're thinking about that too, if under the hood, there are some patterns that emerge which might emulate formal mappings of code segments given enough pre training. I think on the codebase level, that's probably true, but on the method level, the jury is still out.

We’ve released a report on RLFR: Reinforcement Learning from Formally-Defined Rewards for Code Generation (haladir.com/blog/rlfr). The core idea is simple: we present a simple offline and single-turn RL framework that uses formal verification as the primary reward signal for program synthesis in LLMs rather than unit test-based rewards (RLVR), showing improvements over a swath of different coding benchmarks. We fine-tune both Qwen2.5-Coder-7B-Instruct and Qwen3-8B using RLFR and Dafny, a specification language that utilizes Boogie (IVL) and Z3 (SMT solver) to check verification conditions. Despite solely training on Dafny, we show improvements on Python-native coding benchmarks, suggesting generalization past the formal verification domain.

@nabeelqu i don't think its a coincidence that most such cases have later received the caveat that a very similar proof was later found online, i.e, solving erdos by 99% reasoning + 1% web search != 99% web search + 1% reasoning x.com/gdb/status/201…

He put an Erdos Problem into GPT 5.2Pro, no special prompting, and it *thought continuously for 41 minutes and spat out a correct proof*. It's over everyone, we had a good run.

@jibranhutch very cool, for imperative language code gen this seems promising. i'm curious how much the sheer compute available to train frontier models will mimic some formal reasoning via self-reflection even absent explicit RLFR

Say a method has 10 verification conditions extracted from the contracts and invariants, Z3 and Dafny will output the amount proven vs those not proven, we reward that percentage. For our more recent multi-turn setups, it’s 0 and 1 for total correctness with self-play. You can also get pretty creative with weighting certain conditions higher than others, such as post conditions as weighing more than invariants.