Jeraime Griffith

The setup was smooth. I had large files in my project directory that I had forgotten about that was more than the provisioned memory of the vm, hence me listing all file sizes to figure it out.

After some trouble shooting I was able to get a project running @slicervm. In the end I created a bash script that will copy the files I want into the vm and setup @claudeai to run in bypass permission mode.

I can tear it down if my agent goes rouge and quickly spin up another. After a good run I can copy code back to my machine.

@alexellisuk @dansemperepico Sorry thought I shared the link - github.com/jgriffi/comman…

@alexellisuk @dansemperepico The files are .md.txt. You can view them and decide. Would appreciate some feedback. I tend to tweak them every now and then. But generally Claude is good at following these instructions.

You guys all run Claude Code with claude --dangerously-skip-permissions right? Because otherwise how in the world can you sit there accepting every single permission when building something?

@alexellisuk @dansemperepico I use /execute-plan command to kick off the plan (subdivided into tasks). It orchestrates my agents coding-assistant, coding-evaluator in a loop. coding-assistant can use a test-runner agent to write tests. github.com/jgriffi/comman…

@alexellisuk @dansemperepico Happy to

@JeraimeGriffith @dansemperepico Can you share a screenshot? (Redacted if necessary..)

I’m a chemist. I need to say this - because it’s getting dangerous out there. The biggest health myth in the world isn’t about vaccines. Or GMOs. Or fluoride. It’s the root of all of them. It’s called chemophobia - and it’s killing science. Fear of “chemicals” now drives vaccine rejection, GMO bans, food hysteria, and entire political movements. From tampons to tap water, people have been taught to fear chemistry - the very thing that keeps us alive. Chemophobia tells us: “Natural is good.” “Synthetic is bad.” That’s a lie. Botulinum toxin is 100% natural and one of the deadliest molecules known. Aspirin is synthetic and life-saving. We’ve gone from banning harmful substances for good reason…to banning safe, well-tested molecules for emotional reasons. You’ve seen the slogans: “If you can’t pronounce it, don’t eat it.” “Paraben-free.” “Clean beauty.” They sound empowering. But they’re not science - they’re marketing. And they’re making the world dumber, poorer, and sicker. Your body doesn’t care if a molecule comes from a plant or a lab. Vitamin C is vitamin C. Formaldehyde is formaldehyde and your body makes more of it every day than any vaccine ever could. Dose matters. Source doesn’t. This fear isn’t harmless. It shapes public policy. It blocks innovation. It raises food prices. It slows down cancer treatments. Chemophobia is now mainstream and it’s costing lives. Scientists aren’t losing because we’re wrong. We’re losing because fear spreads faster than facts. Because influencers sell fear for clicks. Because lawyers monetize doubt. And because scientists are too tired to fight back. So here’s my message, as a chemist and as a citizen: Learn how toxicology works. Call out chemical fear-mongering. Support policies based on evidence, not emotion. Chemistry isn’t the enemy. It’s the reason you have clean water, safe food, and modern medicine. If we let fear win, we lose all of it.

Flicking through this book at a book shop. Curious that they are no discoveries just north of London, Wales or Cornwall?

@bcherny Wrt to putting up PRs, how do you handle merge conflicts. As I understand it working in git worktrees means each agent gets a copy of the codebase and don’t see what others are doing. As opposed to agent teams where it’s an active collaboration where you can resolve conflicts.

/batch Interactively plan out code migrations, then execute in parallel using dozens of agents. Each agent runs with full isolation using git worktrees, testing its work before putting up a PR. Usage: "/batch migrate src/ from Solid to React"

In the next version of Claude Code.. We're introducing two new Skills: /simplify and /batch. I have been using both daily, and am excited to share them with everyone. Combined, these kills automate much of the work it used to take to (1) shepherd a pull request to production and (2) perform straightforward, parallelizable code migrations.

@HammerToe I’m from a small island in the Caribbean, Montserrat, now living in London. The Caricom region needs this. Logistics and transportation within the region is notoriously difficult. If a simple search can reveal which island has what then we can increase trade. Would love to help.

I didn’t move to Barbados to build a startup. I moved here to slow down. After 30 years in tech - AI at IBM, blockchain at Ripple & Protocol Labs - I wanted a different pace of life. But living here taught me something important about infrastructure...

@ChaiWithJai @hwchase17 @bcherny @trq212 @Vtrivedy10 I use Opus 4.6 and found that in some sessions it quietly downgrades to Sonnet 4.6

@bcherny @trq212 OK, so this is INSANE. 1189 calls to Claude. 100% nerfed down to Sonnet 4.5 in the last 30 days despite Claude Max. I'm so happy I have LangSmith for observability. There could be a bug on how this is reported. But right now, this is really bad... cc: @hwchase17 @Vtrivedy10

@willmcgugan Harlequin is another TUI for SQL

Check out this amazing TUI for SQL databases! 🤓 github.com/Maxteabag/sqlit

The reverse would be great as well. A ⁦@duckdb⁩ plugin or extension for ⁦@neo4j⁩ podcasts.apple.com/gb/podcast/dis…

@grok show me tweets about strigolactones

Quite trippy re-watching Caprica in the age of #LLMs

@pycoders This is great if you are using an LLM to extract data from the json object of an API call. Sometimes you have to ask if all data has been extracted.

A `dict` That Can Report Which Keys Weren't Used peterbe.com/plog/a-python-…

Anyone know of a good open source SQL validator that I can pip install? #DataScientists #SQL

@SGRodriques Would be great to see the untruncated reasoning. A bit hard to follow. It got the correct structure but does the full reasoning lead logically to the molecule it suggestsed.

Finally, the individual reasoning traces that come out of the model are also very compelling. Here it is backing out the chemical structure associated with a natural product. This is a hard task for humans. 4/n

Today we are releasing ether0, our first scientific reasoning model. We trained Mistral 24B with RL on several molecular design tasks in chemistry. Remarkably, we found that LLMs can learn some scientific tasks more much data-efficiently than specialized models trained from scratch on the same data, and can greatly outperform frontier models and humans on those tasks. For at least a subset of scientific classification, regression, and generation problems, post-training LLMs may provide a much more data-efficient approach than traditional machine learning approaches. 1/n