Robert Hoffmann

Tech Lover (Psychology, Marketing, Biotech). Product Owner / R&D Software Engineer (PSPO/PSM certified)̬̤̣̮̩̱̭ Dev, UX, AI. 25+y cross-domain XP

[Robert Hoffmann @itechnologynet]

Found my old CV to be kinda mehh
So made it into an app 😎

https://t.co/XM7G9xqAr9

[Robert Hoffmann @itechnologynet]

@slicknet Yeah, even the max is too limited. I have it.

On light coding tasks I'll use 10% a day (so i actually need 2 max plans to make the month)

On heavy tasks, I'll use 10% per hour ! ...that said, it's only $10/hour, if you compare that to an actual dev.

[Robert Hoffmann @itechnologynet]

@svilendonmez @thsottiaux @sama True, my bad

Still 155h is not 1.5 days 🙃

[Robert Hoffmann @itechnologynet]

@merill @NathanMcNulty Probably something like this

https://t.co/7Zwp4HAJXp

[Burke Holland @burkeholland]

Introducing MAI-Code-1-Flash

Microsoft's latest small coding model.

51.2% on SWE-Bench Pro.

Rolling out now to @GitHub Copilot Free, Pro, Pro+ And Max users.

https://t.co/vQTuCTQTnF

[Robert Hoffmann @itechnologynet]

@brunoborges yeah ..you aint building something like this in auto-mode ^^

100% GPT generated
https://t.co/XM7G9xqAr9

[Robert Hoffmann @itechnologynet]

@MetalGlot @slicknet no, openai and anthropic max plans are still heavily subsidized

and current copilot pricing makes absolutely no sense

https://t.co/XViH2b0tVO

[Robert Hoffmann @itechnologynet]

Dunno WTF they are doing at @Microsoft

Opus 4.5 costs just as much as GPT 5.5

GPT 5.5 was half as expensive as Opus 4.8 before the transition, and now is twice as expensive as Opus 4.8 (4x increase), despite microsoft not paying any royalties to OpenAI anymore 🙄

https://t.co/ifq0b8GjQU

[Robert Hoffmann @itechnologynet]

Dunno WTF they are doing at @Microsoft

Opus 4.5 costs just as much as GPT 5.5

GPT 5.5 was half as expensive as Opus 4.8 before the transition, and now is twice as expensive as Opus 4.8 (4x increase), despite microsoft not paying any royalties to OpenAI anymore 🙄

https://t.co/ifq0b8GjQU

[Robert Hoffmann @itechnologynet]

Dunno WTF they are doing at @Microsoft

Opus 4.5 costs just as much as GPT 5.5

GPT 5.5 was half as expensive as Opus 4.8 before the transition, and now is twice as expensive as Opus 4.8 (4x increase), despite microsoft not paying any royalties to OpenAI anymore 🙄

https://t.co/ifq0b8GjQU

[Robert Hoffmann @itechnologynet]

@pierceboggan @Microsoft @code Before the transition, GPT 5.5 was half as expensive as Opus 4.8

[Robert Hoffmann @itechnologynet]

@pierceboggan @Microsoft @code Oh and all the copilot ppl that were on the annual plan

GPT 5.5, is TWICE as expensive as Oput 4.8

[Robert Hoffmann @itechnologynet]

@pierceboggan @Microsoft @code ..and microsoft doesnt even pay royalties to openai anymore

https://t.co/ItxAIGCw2k

[Robert Hoffmann @itechnologynet]

yeah, i basically commit, run a prompt, scroll through the diffs, and closeouts (i have it write feedback: what this this accompish, how did it make the product better, what could have been done better, what can we do to do better), and commit and move to the next prompt (actually sprint here)

[Robert Hoffmann @itechnologynet]

@loktar00 I mean standalone...

A cli on one screen, and a good IDE on another screen, for tracking what the cli does, actually works pretty well

[Robert Hoffmann @itechnologynet]

Copilot web maybe ..but copilot inside vs code actually worked pretty well, and yoy had access to all the SOTA models at a ridiculous price

... I'd be more worried about ppl using cli's and tui's, and not even tracking their diffs, commits and docs because, well ...cli's and tui's are horrible for that

[Robert Hoffmann @itechnologynet]

@svilendonmez @ServiceHero515 @thsottiaux @sama You said 1.5 days ...but 155h is 6 days

... shady

[Robert Hoffmann @itechnologynet]

@thsottiaux @svilendonmez @sama No way anyone is running a job for 6 days 24/7 on a $20 plan 🤣

...that or GPT was lagging like hell (or hitting the 5h limit every 30 minutes and then sleeping for the rest of the time)

[Robert Hoffmann @itechnologynet]

@PavitraGolchha @mhdcode i stoped using TUI's when win95 came out...

i still use the command line ..but why on earth would i want a tui where it's basically just a flashy command line, when i can have a IDE with real diffs, mermaid diargams and a shitload of cool plugings ?

Norton Commander (1993)

[Robert Hoffmann @itechnologynet]

@tetsuoai @RaminNasibov

[Robert Hoffmann @itechnologynet]

@zebanderson @rezoundous 4x less, same as claude $100 plan

[Sandro @pupposandro]

A few additions to the memory bandiwidth list:

RTX 3060 360GB/s
M4 Max 32 Core Mac Studio 410GB/s
M4 Max 40 Core Mac Studio 546GB/s
Radeon RX 9070 XT 640GB/s
RTX 3080-10GB 760GB/s
M3 Ultra Mac Studio 819GB/s
RTX 3080-12GB 912GB/s
RTX 5080 960GB/s
RTX 6000 960GB/s
RTX 4090 1,008GB/s
Radeon Instinct MI60 1,024GB/s
RTX Pro 6000 1,792GB/s