Update README.md

README.md

@@ -10,8 +10,31 @@ tags:
 - from-scratch
 pipeline_tag: text-generation
 ---
+# 4 April 2026 Update
 
-# LaaLM-v2
+We have some news.
+
+After we thought about how to go with LaaLM, we realized we'd be just spending compute.
+
+We had originally thought to at least deliver it with some fine-tuning of Qwen 2.5 0.5B, but we have decided that LaaLM had enough models.
+
+So we are announcing that we are officially stopping the LaaLM lineage here.
+
+We know it's pretty early with just 2 released models being LaaLM-v1 and LaaLM-exp-v1, and a bunch of failed LaaLM-v2 snapshots that we lost (Like in actual storage sense we don't have acces to the copies.) but we can't see any novel or good perspective about this.
+
+Go try prompting Claude or something like that to act as a Linux terminal, and it'll do much better than LaaLM.
+
+It's actually pretty bad. The model LaaLM-exp-v1 was trained on -Qwen 2.5 3B- literally performed better than the actual LaaLM-exp-v1 when we gave it LaaLM-exp-v1's system prompt.
+
+Or like maybe just install actual Linux.
+
+We are thinking of another new model to continue our adventure in predicting computers with neural networks, but we do not want to say it here to keep some mystery. 
+
+And to make sure we don't announce something that won't make it.
+
+Goodbye.
+
+# LaaLM-v2 (Deprecated. Anything you'll be reading from now on has been either lost or stopped.)
 
 **A small language model trained from scratch to be a Linux terminal. That's it.**
 
@@ -142,7 +165,7 @@ We haven't tried the new data generator because of budget problems.
 
 But we thought in the meantime that Transformers are too heavy for what LaaLM does.
 
-So we are going to use a LSTM based architecture.
+So we are going to use an LSTM-based architecture.
 
 LSTMs are not designed for big models. But LaaLM is very simple, so an LSTM is better for it.
 
@@ -180,7 +203,7 @@ Our reason is that we have better projects to spend our compute on than a bash p
 
 LaaLM has definitely been a fun experience for us, but we can't just spend precious compute for something this experimental and non-useful.
 
-Maybe future models will keep coming but we definitely recommend you can stop expecting new models after LaaLM-v2.
+Maybe future models will keep coming, but we definitely recommend that you may stop expecting new models after LaaLM-v2. (Well this turned out to be correct.)
 
 ---
 
@@ -242,9 +265,7 @@ We thought of basic bash, but we don't know if we can handle it.
 
 ## Status
 
-Currently training. If you're seeing this, the model might not be uploaded yet or might change significantly.
-
-Check back later or watch the repo for updates.
+Deprecated.
 
 ---