nanomader

707 posts

nanomader

@nanomader

building @getgoalcoder

Entrou em Ağustos 2020

228 Seguindo30 Seguidores

nanomader@nanomader·4h

@mattyp

QME

171

matt palmer@mattyp·1d

caffeinate -i

roon@tszzl

people are walking around with their laptops slightly ajar to keep their agents running

Deutsch

388

65.2K

nanomader@nanomader·1d

I am building GoalCoder, an agent orchestrator for bounded, verifiable changes to existing code repositories. It will be closed source. That gives me a duty to be specific about network access, provider use, local artifacts, logs, and what users can and cannot verify

English

nanomader@nanomader·3d

ummm

nanomader@nanomader·4d

Bot Wars 2026

Chris@chatgpt21

I literally just watched GPT-5.5 via codex beat an Amazon customer associate in real time. 💀 I asked it to get me a refund, and I watched it navigate the settings, cancel the subscription, then it went step further into the help page. I thought it was going to request a phone call (which would prompt me to take over) Instead, it opened: “Chat with an associate now.” That’s when I sat up on my couch because I knew it was going to get real The agent said: “Your subscription is active.” And GPT-5.5 immediately explained that it only shows as active because cancellation leaves access through the billing period, but that I wanted it stopped now and refunded. And my jaw just hung open, it was the first time I watched sand handle a customer service agent for me in real time Once the agent confirmed the refund, it just ended the chat no mercy no thank you LMAO First time I’ve watched a human customer service agent get outmaneuvered by AI in real time. And it made me 15$! almost paid for itself in 5 minutes

English

nanomader@nanomader·4d

@kskrygan Well, finally, JetBrains is waking up. Just do it

English

Kirill Skrygan@kskrygan·5d

Would you be interested if JetBrains releases a totally local AI agent, working 100% on your laptop, using our code insight engine and deeply integrated into the IDE? Yes, it will be probably 1 month behind the very recent frontier models, but no token blood bath anymore WDYT?

English

809

235

7.2K

485.9K

nanomader@nanomader·5d

@daniel_mac8 @sama 5.5-codex would be nice, or 5.5-codex-max if they feeling ambitious

English

Dan McAteer@daniel_mac8·5d

@sama release 5.6?

English

2.5K

Sam Altman@sama·5d

wow y'all love 5.5 we should think of something nice to do to celebrate!

English

2.6K

290

11.2K

772.1K

nanomader@nanomader·5d

@haider1 This speed is unbelievable. Just a few days ago 5.5-codex was spotted. Usually they stay very quiet about new models until actual release

English

1.1K

Haider.@haider1·5d

openai is already teasing with gpt-5.6 so if you checked the internal codex logs to verify which model was being used, and you found the rollout mapping most calls were routed to gpt-5.5, but one entry appears to show gpt-5.6, which means the codex environment may have had access to a rollout entry labeled

English

1.3K

236.4K

nanomader@nanomader·6d

AI has made software dramatically easier to create but it has not made software easier to trust. Sooner or later, we need public, verifiable proof that certain AI review prompts were actually run, against specific artifacts, with specific outputs. I mean real receipts, not "trust me"

English

nanomader@nanomader·6d

@PavelSnajdr @elliotarledge rolling out powerful model doesn't mean people automagically will be out of jobs. Adoption rate is a thing

English

Pavel Snajdr@PavelSnajdr·6d

@elliotarledge open your eyes and stop expecting what obviously isn't coming they could have rolled out GPT Pro for 20x cheaper than they did, people could have been out of jobs already did they? instead they raised the price

English

956

Elliot Arledge@elliotarledge·6d

if ~gpt-5.7-spark has vision and is roughly gpt 5.5 medium levels, i think a lot of white collar jobs could be replaced. its up to openai to make the computer use rl environments robust enough, and to get the training hparams just right. very impressive and can do a lot for me right now, just needs speed and polishing.

English

216

16.3K

nanomader@nanomader·6d

Yes, it works but it's not recommended for many reasons. Please avoid doing it unless you do not care about output quality. To make it clear: switching reasoning models within discussion is fine, switching LLM models within discussion is not recommended and it's better to start new thread

English

James Patton@jpattondesign·6d

@nanomader @JakeKAllDay @thsottiaux …is this not something you should do? Works fine for me

English

Tibo@thsottiaux·26 Nis

Looking at the traffic dashboard for Codex just now, it would be scary if we didn't have a lot more compute coming online in the coming weeks. All according to plan fortunately.

English

251

101

4.9K

195.6K

nanomader@nanomader·26 Nis

@JakeKAllDay @thsottiaux I hope you are not toggling in the middle of discussion between different models

English

150

Jake@JakeKAllDay·26 Nis

@thsottiaux @thsottiaux how about an auto mode ala cursor so I don’t have to keep toggling bw 5.4 mini high, 5.5 low, 5.5 med, etc? Save you some compute with efficient routing…

English

4.1K

nanomader@nanomader·26 Nis

this + "Codex Review: Didn't find any major issues. 👍" is the best feeling. Local code review works great!

English

nanomader@nanomader·26 Nis

@mikeysee Yes, good prompt. For the bugs I use following one: While we are here with current context, let's do review of nearby surrounding code and steps to see perhaps is there anything of value we could think and possibly upgrade and/or fix.

English

Mikeysee@mikeysee·26 Nis

This has been a big unlock for me lately: "Is there anything else we should consider?" Use this magical phase frequently. It really helps with the sycophancy and over adherence to the initial plan once implementation starts. It works really well after task implementation and befre PR submission to make sure you have considered all the edges.

English

2.3K

nanomader@nanomader·26 Nis

@dosco I have seen both sides of this in my 9-5 and sadly I do not have recommendation as to how to help people make this transition, because it requires having a open mind and ability to ask a lot of questions

English

spacy@dosco·26 Nis

if your whole job is hand writing code even code that you think is fairly complex. i highly suggest taking a week off to get into LLMs. this message might not find the right audience on here since we're all pretty AI pilled but people out there need to hear this.

English

713

nanomader@nanomader·26 Nis

@willdepue @elder_plinius

QAM

will depue@willdepue·26 Nis

AHHHHHH WHY IS THE SANDBOX NETWORK BLOCKED AHHHH

will depue@willdepue

yes local full-access coding agent was always the right interface, but it’s worth noting that the reason that took off so hard is how fucking unbelievably bad code interpreter was executed. tried today: cant download packages, dies and wipes itself, errors. massive unforced error

English

16.4K

nanomader@nanomader·26 Nis

@de_zolaa @RaminNasibov I fully agree because I am one of such person even though I spend ±15 hours daily in front of the computer working with AI. I do not have source of this image, I saved it around 10 years ago from Reddit (it's not mine)

English

Aura Fårmer@de_zolaa·26 Nis

@nanomader @RaminNasibov You know.m these kinda people do not need any distraction... They really have a peaceful mind

English

Ramin Nasibov@RaminNasibov·25 Nis

I saw a guy at coffeeshop today. No iPhone. No laptop. No tablet. Just sitting there. Drinking his coffee.

English

2.4K

3.3K

54.4K

1.8M

nanomader@nanomader·26 Nis

We are getting closer to the level of Bosch "The Garden of Earthly Delights" detail level with chatgpt images v2

English

nanomader@nanomader·25 Nis

@VictorInFocus @gdb beautiful

English

V@VictorInFocus·25 Nis

@gdb Building a character consistency tool: same character, different scenes. GPT-5.5 got the app together fast, and GPT Image 2.0 is making the actual swaps work