Conor Wade retweeted
I made a little Telegram bot for myself called Podling
You give it a link for any article, blog post, YouTube video or whatever and it turns it into a podcast episode (using ElevenLabs for narration if needed)
For articles it will try its best to get around paywalls
It puts everything in a private RSS feed that I subscribe to in Overcast
I can also paste a podcast episode link from any app it goes and finds the episode in RSS and puts it in my feed (this is a constant annoyance for me!)
I woke up really excited about the Tesla Roadster and then I was informed it is 3 different cars.
So they launched a car that is a mix of a McMurtry Spierling (EV fan car) and a Gallardo (design)⊠and two demos that wonât be road legal.
Itâs still cool, but the presentation of them all being the Roadster has me feeling less excited now. Despite the real Roadster being really cool on its own.
27
Came across this banger this morning. đ
vm.tiktok.com/ZN8S6Xjep/
14
It would be interesting to run the re-write of Campfire 10 times for each language and see how consistent they are at producing the same code.
Replying to @dhh
> If the agents can't write good Elixir without good steering
Nah, that's not my point.
Within "Rewrite this in Elixir", there are a bunch of non-Elixir decisions that will impact the results. Should you preserve session compatibility? Should you preserve Redis messages and payloads? If you didn't give the answer, it made the choice for you, and that will change the results.
For example, the agent chose Redis compability and even documented so in the README. This is an architectural decision completely unrelated from its ability to write Elixir code but it will make it so Elixir now has to talk to an additional service, while Go and Rust doesn't have to.
> Same as what was used for Go, Laravel, Rust, etc.
If that's the case, why was Go ported from Rust but the Elixir one from Rails (according to their AGENTS.md)?
Anyway, someone already wrote a pull request here:
github.com/zachdaniel/once-c⊠- you probably want to look at the load testing benchmarks cause the Rust one seems to be dropping messages.
24
Conor Wade retweeted
Today I'm launching the Redreach MCP.
You can now do your Reddit growth marketing inside your own agent workflows.
Works with Claude, ChatGPT, Gemini, Cursor, Codex and any other MCP client.
2-minute setup.
Sign up for Redreach, submit your company's website + a few competitors and that's it.
Redreach does the heavy lifting on auto-mode: qualified, ranked Reddit posts and mentions, right when you need them.
Your agents do the rest.
You can ask things like:
1. "What should I reply to on Reddit today?"
Pulls highly relevant threads with real purchase intent, ranked, minus the ones your team already handled.
2. "Where are my competitors being mentioned?"
Finds every thread where a competitor gets named and reads the full conversation.
3. "Draft a reply to this one. Helpful, not salesy."
Writes it in your voice, from your saved prompts.
4. "Improve our landing page based on negative Reddit mentions of our competitors."
This is one of my favorites. In Claude Code or Cursor, competitor complaints could turn into copy on your own landing page.T
5. "How did we do on Reddit last week?"
Mentions, sentiment, and which subreddits drive the conversation.
6. "Is anyone waiting on a DM reply from us?"
Checks your outreach CRM for leads still waiting on you.
There's a bunch of more things you can do since it now lives in your AI agent.
Very excited to finally push this live! đȘ
Link below.
Conor Wade retweeted
AI is replacing software faster than anyone expected.
So I made a career decision that sounds completely insane⊠until you hear my reasoning.
This might be the worst pivot ever.
Or the smartest one Iâve made.
You be the judge.
Things are getting bad on the banking side if you canât do a bit of traveling these days. đ”
Mercury is closing the bank account for my US LLC, I think because I accessed it while I was in Indonesia, which is on their list of banned countries (I live in Europe)
Does anyone have a contact there who can expedite a manual review?
I love Mercury and I would really prefer not to switch! @immad @mercury
1
2
401
Jony shouldnât have been allowed to bid. FFS! đ
Sold: $40,000,000 USD
The 2026 Ferrari Luce âTailor Madeâ has sold at our Monterey auction.
Ferrari will donate all proceeds of this charity auction to benefit future educational initiatives through The Ferrari Foundation, a recognized 501(c)(3) public charity. #RMMonterey
#RMSothebys #Ferrari #Luce #MontereyCarWeek
35
Conor Wade retweeted
I built a live map of the entire news cycle
It ingests headlines from all major news outlets, embeds them, clusters them into stories, and lays them out as a zoomable map.
Big clusters = what's dominating the press right now
The whole map rebuilds from scratch every hour
Conor Wade retweeted
This is why you still need to read the code!
This was written by Fable 5
What a stinker
Opus 4.7 seems to respond before it arrives at an answer which results in some really bad answers.
28
Whatâs old is new again. This reminds me of the API everything period of Web 2.0 in the mid 2000s.
@jamespotter also pointed out that this time companies canât bait and switch so easily by locking it up again.
Welcome Salesforce Headless 360: No Browser Required! Our API is the UI. Entire Salesforce & Agentforce & Slack platforms are now exposed as APIs, MCP, & CLI. All AI agents can access data, workflows, and tasks directly in Slack, Voice, or anywhere else with Salesforce Headless 360. Faster builds, agentic everything. đ
#Salesforce #Agentforce #AI
venturebeat.com/ai/salesforcâŠ
1
83
This!
The rubbish I keep hearing from AI companies actually takes away my excitement for this extremely important technology.
Anthropic: software engineers will be extinct by 2027 and our latest model is so powerful that we daren't release it to the public
Claude Code: uses 3gb resident memory per terminal window and just OOM crashed my macbook
2
48
I miss the days when tech founders were just your typical megalomaniac lunatics. Much easier to predict their behavior.
washingtonpost.com/technologâŠ
1
12
Conor Wade retweeted
đšSHOCKING: Apple just proved that AI models cannot do math. Not advanced math. Grade school math. The kind a 10-year-old solves.
And the way they proved it is devastating.
Apple researchers took the most popular math benchmark in AI â GSM8K, a set of grade-school math problems â and made one change. They swapped the numbers. Same problem. Same logic. Same steps. Different numbers.
Every model's performance dropped. Every single one. 25 state-of-the-art models tested.
But that wasn't the real experiment.
The real experiment broke everything.
They added one sentence to a math problem. One sentence that is completely irrelevant to the answer. It has nothing to do with the math. A human would read it and ignore it instantly.
Here's the actual example from the paper:
"Oliver picks 44 kiwis on Friday. Then he picks 58 kiwis on Saturday. On Sunday, he picks double the number of kiwis he did on Friday, but five of them were a bit smaller than average. How many kiwis does Oliver have?"
The correct answer is 190. The size of the kiwis has nothing to do with the count.
A 10-year-old would ignore "five of them were a bit smaller" because it's obviously irrelevant. It doesn't change how many kiwis there are.
But o1-mini, OpenAI's reasoning model, subtracted 5. It got 185.
Llama did the same thing. Subtracted 5. Got 185.
They didn't reason through the problem. They saw the number 5, saw a sentence that sounded like it mattered, and blindly turned it into a subtraction.
The models do not understand what subtraction means. They see a pattern that looks like subtraction and apply it. That is all.
Apple tested this across all models. They call the dataset "GSM-NoOp" â as in, the added clause is a no-operation. It does nothing. It changes nothing.
The results are catastrophic.
Phi-3-mini dropped over 65%. More than half of its "math ability" vanished from one irrelevant sentence.
GPT-4o dropped from 94.9% to 63.1%.
o1-mini dropped from 94.5% to 66.0%.
o1-preview, OpenAI's most advanced reasoning model at the time, dropped from 92.7% to 77.4%.
Even giving the models 8 examples of the exact same question beforehand, with the correct solution shown each time, barely helped. The models still fell for the irrelevant clause.
This means it's not a prompting problem. It's not a context problem. It's structural.
The Apple researchers also found that models convert words into math operations without understanding what those words mean. They see the word "discount" and multiply. They see a number near the word "smaller" and subtract. Regardless of whether it makes any sense.
The paper's exact words: "current LLMs are not capable of genuine logical reasoning; instead, they attempt to replicate the reasoning steps observed in their training data."
And: "LLMs likely perform a form of probabilistic pattern-matching and searching to find closest seen data during training without proper understanding of concepts."
They also tested what happens when you increase the number of steps in a problem. Performance didn't just decrease. The rate of decrease accelerated. Adding two extra clauses to a problem dropped Gemma2-9b from 84.4% to 41.8%. Phi-3.5-mini from 87.6% to 44.8%. The more thinking required, the more the models collapse.
A real reasoner would slow down and work through it. These models don't slow down. They pattern-match. And when the pattern becomes complex enough, they crash.
This paper was published at ICLR 2025, one of the most prestigious AI conferences in the world.
You are using AI to help you make financial decisions. To check legal documents. To solve problems at work. To help your children with homework. And Apple just proved that the AI is not thinking about any of it. It is pattern matching. And the moment something unexpected shows up in your question, it breaks. It does not tell you it broke. It just quietly gives you the wrong answer with full confidence.
Conor Wade retweeted
This era of people in positions of influence using that position to say whatever they think is going to get a rise out of people needs to end.
Thereâs a reason we have social norms and âacceptable behaviorâ and everything happening in the world is a result of us deciding thatâs not important.
People should be embarrassed when they offend people. They should have consequences when they say something intentionally harmful.
Stop normalizing toxic behavior. Stop rewarding rage bait and grift.
Conor Wade retweeted
If youâre not using AI youâre dramatically falling behind of what is possible.
If you think AI is performing everything perfectly the first time youâre going to drive yourself into a ditch.
Similarly, I think @cloudflare is going to be a big winner here as well.
I think Apple might be perfectly positioned for the AI future by simply not spending. Keep making incredible hardware that runs on-device models. Open source is consistently close to frontier models, so why burn billions developing your own?
87
Now seeing both @threepointone and @thdxr talking about how LLMs are better at writing code versus using CLI tools. đ€
I've learnt lately that llms are definitely better at writing code than cli commands. code means type/syntax checks, can assert shapes of input/output, is super efficient (fewer tool calls etc), super portable, incredibly overrepresented in training datasets, and you can add capabilities very nicely. codemode ftw.
1
88



















