Martin McFly

Debilitating caffeine dependancy ☕️

[Martin McFly @martinmark]

@qualtrim Dual axis charts. NOPE.

[Martin McFly @martinmark]

@hiddensmallcaps 1. Neo DCs are better 2. TPUs are a reach to not be locked into Nvidia's allocations (if they have better economics, why offer Nvidia at all?) 3. They're struggling to get memory allocations for TPUs - they're new, they have purchasing power but not the trust of memory producers

[Martin McFly @martinmark]

@jplej1337 @PsudoMike Hopefully competition and incentives ✊

[Martin McFly @martinmark]

@encrypted I hearted it, but I hate it

[Martin McFly @martinmark]

@ansonlin Scotia Amex Gold (money with them / fee waived), Rogers MC (Costco 🤷‍♂️), WS Visa (everything else)

[Martin McFly @martinmark]

@nigkang @daniel_mac8 "3x 1200w PSU" - yeah, my old-ass house just burnt down

[Martin McFly @martinmark]

@charlieholtz Yeah 5.5s definitely better. For sure felt the difference, and felt like the Codex harness may have been doing the improvements. ✊Glad you caught it, back to running my life in parallel

[Martin McFly @martinmark]

@LottoLabs There were a lot of 3090 pre-builds last year. Toronto area. Got some sub-$1400 4090s right now as well. Gotta make sure your breakers can support the draw though, mine couldn't at load 😭

[Martin McFly @martinmark]

@LottoLabs The price/perf for the 395 Max+ is really good. Grabbed one last year before memory went bananas

[Martin McFly @martinmark]

@AvgDatabaseCEO But they gave you such good feedback and all that stress testing. They need a recurring 80% discount, at least - it's only fair

[Martin McFly @martinmark]

@openclaw “Held together with hope” resonates for me - I’ll give it a try

[Martin McFly @martinmark]

@valigo We really gotta start reading the ToS

[Martin McFly @martinmark]

@RhysSullivan womp womp. This used to be fine https://t.co/gVgQQhvxDf

[Martin McFly @martinmark]

@pontusab Or too early if there's no pagination ;)

[Martin McFly @martinmark]

@venturetwins If you don't smell at the end of a hackathon, you're not doing it right

[Martin McFly @martinmark]

@thorstenball @lewis_b_metcalf The main value I got from this was a better understanding of Thread usage in Amp 👍 My workflow's similar, but with some manual shenanigans in CC. Gonna have to have another look at Amp

[Martin McFly @martinmark]

@kitlangton I like this a lot - "we can try to trust the docstring... and pray it's up to date" 😂

[Martin McFly @martinmark]

@colin_fraser Standard JSON fails to amaze me anytime I want to add a comment or am scrolling a file to find where the trailing comma is

[Martin McFly @martinmark]

@TriDeapthBear Would love some evals on this. In LLM prompting passing nested JSON as completion content can create noise for the models. There have been some studies that they're better at kv, markdown, and XML vs. JSON. key:value make sense, but I find these structured blobs questionable

[Martin McFly @martinmark]

@notnotstorm Would it be easier to use the Github integration after the issues instead of the worktrees?