SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

Tech Giants Quietly Rehire Developers After AI Coding Experiment Backfires

0
·1 views

Major tech companies including Google, Amazon, and Meta laid off tens of thousands of developers in 2024, betting that AI tools could replace human coders and cut costs. However, real-world deployments revealed serious problems, including AI-generated code containing up to 1.7 times more errors than human-written code and experienced engineers becoming 19% less productive when using AI assistance. Amazon reportedly suffered four critical system errors in 90 days linked to AI-assisted code changes, highlighting the technology's struggle with complex, context-dependent software environments. Gartner now projects that 50% of companies that eliminated roles due to AI will rehire for those same positions by 2027, a trend analysts are calling 'boomerang hiring.' The reversal underscores that while AI can generate code rapidly, it lacks the business context understanding and judgment required for reliable, large-scale software development.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Kolaboratr Launches as a Collaboration Platform for Indie Creators and Project Leads

Vasken Interactive Studios LLC has launched Kolaboratr, a creator collaboration platform designed to connect developers, artists, writers, designers, composers, and other creatives with projects and teams. Project leads can post listings called 'Kolabs,' detailing their project, required roles, and compensation arrangements, while creators build profiles showcasing their skills, tools, and portfolios. The platform is in its early stages and is actively seeking its first wave of users to help establish the community. Kolaboratr also features a built-in tool called Kontracts, which allows collaborators to formally document key project terms such as ownership, revenue sharing, credit, and exit conditions. The studio clarifies that Kontracts are not a substitute for legal counsel but aim to replace informal agreements scattered across chat messages and DMs.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

7 AWS Resources That Can Silently Inflate Your Cloud Bill Over Time

A technical review highlights seven commonly overlooked AWS resources that can accumulate unnecessary costs over time, including idle EC2 instances, unattached EBS volumes, and forgotten S3 data. Automated processes such as snapshot creation and multipart uploads can quietly grow storage usage without defined retention or cleanup policies. Underutilized EC2 instances are flagged as a key area to investigate, with AWS Compute Optimizer recommended for identifying rightsizing opportunities. The guide also cautions against hasty deletions, advising teams to first understand dependencies and workload requirements before removing any resource. Reviewing storage classes, lifecycle policies, and Elastic IP allocations are among the additional steps suggested to bring cloud spending under intentional control.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

12-Year-Old Developer Rebuilds Failed Hackathon as Standalone Platform After Zero Submissions

A 12-year-old developer from Tamil Nadu launched KODA Code Jam, an online coding tournament, but received zero submissions in its first run. He identified the problem as excessive friction — users had to navigate five steps inside an unrelated chat app just to submit an entry. In response, he rebuilt the Jam as a dedicated standalone platform with a clean landing page, a simple submission form, a public showcase wall, and spam-resistant email voting. The revamped platform cross-promotes his broader ecosystem, which includes an AI mentor tool called KODA, a youth coding community called DOJO FEED, and a learning platform under development. Winners receive in-app rewards including expanded AI memory, a special badge, and a permanent place in a Hall of Champions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Q4_K_M Outperforms OpenAI's MXFP4 by 1.8x in Local LLM Speed Test on Apple M2

A developer ran a head-to-head speed test comparing two local LLM quantization formats — Q4_K_M and MXFP4 — on the same Apple M2 MacBook with 24GB unified memory using Ollama. Alibaba's Qwen3-14B in Q4_K_M format averaged 4.7 tokens per second, completing a 200-token task in 44 seconds, while OpenAI's gpt-oss-20B in MXFP4 managed only 2.6 tokens per second, taking nearly 71 seconds for the same workload. Despite MXFP4 being marketed as a faster, low-latency format optimized for local inference, it was approximately 1.8 times slower in warm-run trials. The tester theorizes that the M2 chip lacks native FP4 hardware acceleration — a feature introduced only in the M4 — forcing MXFP4 weights to be dequantized to FP16 before processing, negating its theoretical advantages. The results suggest that MXFP4's performance benefits may be hardware-dependent and not yet fully realized on older Apple Silicon.

0 comments Read more at DEV Community