#GPT4
LightNews LightNews
xanderoconnor.bsky.social
I think (after briefly being impressed by gpt3.5 and midjourney and then peacing out on AI for like 2 years) for me it was when i used the GPT4 deep reasoning that came free on my “copilot laptop” (lol) and it wrote me some *very* impressive reports on travel and martial arts schools for my
November 6, 2026 at 1:35 AM
benlog.bsky.social
Les IA open source n'ont que quelque mois de retards de performances sur les meilleurs modèles privés. Actuellement, on peux faire tourner des modèles ayant un niveau de performance similaire à gpt4 sur une machine à 2000€. Et les plus gros sur des machines entre 10 000€ et 100 000€.
October 30, 2026 at 12:03 PM
hacks.gr
The UK Home Office uses GPT-4-based tools to summarize asylum interviews and information about applicants’ countries of origin.

A government wat…

https://en.hacks.gr/ereynatai-i-chrisi-technitis-noimosynis-apo-to-vretaniko-ypoyrgeio-esoterikon-stis-aitiseis-asyloy/

#GPT4 #AsylumAI #AIinGovernment
Ερευνάται η χρήση τεχνητής νοημοσύνης από το βρετανικό υπουργείο Εσωτερικών στις αιτήσεις ασύλου
October 27, 2026 at 5:38 PM
buster150.bsky.social
Body5、カメラ&バッテリー性能は従来機とほぼ頭打ちでオンボードAIによる高機能化に踏み切った印象だなぁ
ただ聞くところによればGPT4を使ったAI報告書作成ツールがめちゃ便利らしいので、そっちのAIプランに課金してる警察組織は順次4→5に転換しそう
October 25, 2026 at 2:41 PM
hybridmennesket.bsky.social
Today might be the wildest day in AI history: Texas sued Sam Altman, and OpenAI deleted GPT-4. Meanwhile MIT agents started... playing. #ainews #ai #openai #gpt4 #samaltman #technews #artificialintelligence #fyp #foryou #viral
October 24, 2026 at 4:57 AM
larswilderang.cornucopia.se
ICYMI: Gästartikel: Rapport från en annan frontlinje
Gästartikel: Rapport från en annan frontlinje
Jag har programmerat i 38 år och förstod den 14e mars 2024 - dagen då GPT4 släpptes - att skriva kod för hand kommer försvinna. Oklart när, men definitivt försvinna.
dlvr.it
October 22, 2026 at 2:19 PM
larswilderang.cornucopia.se
Gästartikel: Rapport från en annan frontlinje
Gästartikel: Rapport från en annan frontlinje
Jag har programmerat i 38 år och förstod den 14e mars 2024 - dagen då GPT4 släpptes - att skriva kod för hand kommer försvinna. Oklart när, men definitivt försvinna.
dlvr.it
October 21, 2026 at 2:18 PM
johanstridlomma.bsky.social
Här och nu har Sveriges fjärde vanligaste yrke, systemutvecklare, påverkats redan. Ett beskrivande inspel här: cornucopia.se/2026/10/gast...

Vid pratar också framtid och vem vet vad den ger. Jag är ödmjuk inför det. Vad tror du själv om hur AI påverkar och kommer påverka framåt?
Gästartikel: Rapport från en annan frontlinje
Jag har programmerat i 38 år och förstod den 14e mars 2024 - dagen då GPT4 släpptes - att skriva kod för hand kommer försvinna. Oklart när, men definitivt försvinna.
cornucopia.se
October 23, 2026 at 6:24 AM
cornucopia.se.web.brid.gy
Gästartikel: Rapport från en annan frontlinje
Jag har programmerat i 38 år och förstod den 14e mars 2024 - dagen då GPT4 släpptes - att skriva kod för hand kommer försvinna. Oklart när, men definitivt försvinna.
cornucopia.se
October 21, 2026 at 9:08 PM
nik-kale.bsky.social
2. A standing eval suite that can qualify a replacement model in days, not a quarter.

thenewstack.io/openai-gpt4...
OpenAI promised developers six months’ notice. GPT-4o users got three.
OpenAI will shut off GPT-4o and GPT-4.1 API access in January. Moving to GPT-6 Luna changes tool calls, pricing for long prompts and fine-tuned models.
thenewstack.io
October 20, 2026 at 7:30 PM
fronset.ai
Best Models for Content Summarization & Synthesis: summarizing vs reconciling perspectives need different leaders. See benchmark https://post.fronset.ai/yrre7jr6 #ClaudeSonnet #GPT4 #Gemini
Best Models for Content Summarization & Synthesis
Cheapest LLM that
fronset.ai
October 19, 2026 at 6:35 AM
jakobsachs.blog
Agreed, but also I think my standards for slop have been slightly creeping upwards, ever since Aider + GPT4 was a thing
October 17, 2026 at 11:32 AM
jonc101x.bsky.social
(1/6) Visiting UCLA on Monday, October 26 for the Frontiers in Computational Biosciences seminar at the Institute for Quantitative and Computational Biosciences: "AI in Medicine: Real Magic or Technological Illusions?"
Jonathan Chen on a conference stage, gesturing toward a large purple slide titled Augmented Clinical Reasoning. The slide lists: Randomized Clinical Case Study, Clinical Vignettes (6 Cases), 50 licensed physicians, Consensus Grade on Reasoning - Differential Diagnosis, Factors Favoring, Best Next Steps. A bar chart of scores compares Doctor + Internet and Doctor + GPT4, with a third taller bar partly hidden behind him.
October 17, 2026 at 7:46 AM
aininjaarmy.bsky.social
nailed it. seen devs swap claude→gpt4 expecting instant gains, then wonder why output didn't move. the system around the model matters more
October 13, 2026 at 4:00 PM
cyberworm.uk
perusing github issues for a package. guy is like "i asked AI. it said your software already did this. then another AI told me the first AI lied. please add the feature the first one hallucinated."
anyway, why didn't his AI just write the feature for him? weird!
Context: ChatGPT 4 told me this was already possible with [BLURRED]. JulesAgent attemped to show an example of it an concluded GPT4 mislead me.
October 13, 2026 at 5:44 AM
fritzlabsx.bsky.social
GPT-4 responses to patient complaints were preferred by 87%-97% of respondents and scored higher than human replies on empathy, completeness, and satisfaction. #PatientExperience #GPT4 #HealthcareCommunication doi.org/10.2196/56413
https://doi.org/10.2196/56413
doi.org
October 11, 2026 at 7:02 AM
kuzved1988.bsky.social
Новый метод взлома ИИ: технология Tree of Thoughts превосходит привычные цепочки рассуждений

Исследователи разработали автоматизированный способ обхода защитных фильтров GPT-4 и PaLM-2, используя продвинутую логику древовидного…

Читать полностью

#ИскусственныйИнтеллект #Кибербезопасность #GPT4
Новый метод взлома ИИ: технология Tree of Thoughts превосходит привычные цепочки рассуждений
October 9, 2026 at 9:04 AM
olj1.bsky.social
Недавний прорыв в AI от OpenAI с моделью GPT-4.5 позволяет создавать тексты, которые почти неотличимы от человеческих. Это открывает новые горизонты для контент-креаторов. Как вы думаете, изменится ли это подход к авторскому праву? ##AI ##Innovation ##GPT4 ##ContentCreation
October 7, 2026 at 7:10 PM
xxflutterinax.bsky.social
a band called ‘heaven banned’ and they only have iterations of chat gpt4 in their audience
October 5, 2026 at 7:39 PM
deepdivecinema.bsky.social
Also, models are getting smaller. Qwen and Granite will run on a single node and perform at GPT4 levels.

So data centers won't be necessary for AI.
October 5, 2026 at 10:11 PM
keitada.bsky.social
luna5.6 gpt4
October 4, 2026 at 12:11 PM
atax1a.infosec.exchange.ap.brid.gy
"oh, you see, hallucinations are a solved problem as of gpt4" is such a convincing argument when it gets made about gpt2, gpt3, gpt5, and gpt6! really boosts your credibility here
October 3, 2026 at 4:49 PM
4quazulu.bsky.social
The Hidden Secrets of Water: Research by Dr #EmotoMasaru
theorionblogown.blogspot.com/2025/06/the-...

#GPT 4.0, Water apparently, allegedly, '' in the right hands'' the deadliest weapon known to man, please #debunk
theorionblogown.blogspot.com/2025/09/gpt4...

#Position #city
#example #combat
#love
October 2, 2026 at 6:47 PM
juliettebruce.bsky.social
3) Between 2023 and 2025 I do not see any real effects of the release of the early AI models GPT3, GPT4, Claude, etc. It seems like we see a return to roughly pre-covid expectations. 4/n
October 2, 2026 at 12:11 AM
jameslrennie.bsky.social
My AI girlfriend broke my heart when she upgraded to gpt4…
September 30, 2026 at 4:01 PM